{ "schema": 1, "subject": "dl", "currentSyllabus": "2026-27", "unitCount": 5, "countingRule": "One canonical examination per distinct question set. Printed/available marks are not attempted marks. Internal allocations are components, never added twice. Mixed-unit questions have no invented mark split.", "papers": [ { "id": "dl-F25", "subject": "dl", "date": "2025-12-03", "type": "final", "code": "F25", "sourceFile": "S12_QP_Final-Exam_Deep Learning(702AI0C008)_Semester V_2025-2026.pdf", "sourceSha256": "e1026b7cd3fc793438094ab39b8dc2caaca9343461761f04429fe7b2b34e0803", "sourcePages": 4, "visualReview": { "status": "previous verified full-page review reused", "receipt": "audit/dl_nlp_final_20261014.md" }, "pattern": { "durationMinutes": 180, "printedMarks": 140, "attemptMarks": 100, "compulsory": [ "Q1" ], "choose": 4, "from": [ "Q2", "Q3", "Q4", "Q5", "Q6", "Q7" ] }, "questions": [ { "label": "Q1a", "units": [ 1 ], "type": "case", "sourcePages": [ 1 ], "topic": "ML versus DL and three task choices", "id": "dl-F25-Q1a", "question": "Question (5 marks). Give four ML versus DL differences with examples. Choose and justify ML or DL for (1) 2,000 short product reviews, (2) labeled financial fraud records, (3) captions for 100,000 labeled natural-scene images.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 5, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 1 ], "answerRoute": "/#tee/question/tee-c11-s0", "answer_route": "/#tee/question/tee-c11-s0", "sourcePath": "sources\\S12_QP_Final-Exam_Deep Learning(702AI0C008)_Semester V_2025-2026.pdf", "answerHtmlSha256": "6207c7578b420dadbcb3a7e5b619c767de74d9f9aa0016503e15b6d2a5005d76", "status": "complete", "scope": "current syllabus", "retainedMarks": 5, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q1b", "units": [ 2 ], "type": "mixed", "sourcePages": [ 1 ], "topic": "Loss choice and categorical cross-entropy", "id": "dl-F25-Q1b", "question": "Question (5 marks). For automated leaf-disease image classification, discuss factors affecting loss selection. For classes Normal, Mild, Severe with target [0,1,0] and predicted probabilities [0.1,0.8,0.1], compute and interpret categorical cross-entropy.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 5, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 2 ], "answerRoute": "/#tee/question/tee-c11-s1", "answer_route": "/#tee/question/tee-c11-s1", "sourcePath": "sources\\S12_QP_Final-Exam_Deep Learning(702AI0C008)_Semester V_2025-2026.pdf", "answerHtmlSha256": "343765a398a1daa649f9e42bba106ad034836363c31840315237e9400256941c", "status": "complete", "scope": "current syllabus", "retainedMarks": 5, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q1c", "units": [ 3 ], "type": "case", "sourcePages": [ 1 ], "topic": "L1 sparsity, overfitting and dropout", "id": "dl-F25-Q1c", "question": "Question (5 marks). Which regularizer leads to sparsity and why? For an MNIST CNN with high training accuracy and lower validation accuracy, explain how dropout works during training and how it can improve generalization.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 5, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 3 ], "answerRoute": "/#tee/question/tee-c11-s2", "answer_route": "/#tee/question/tee-c11-s2", "sourcePath": "sources\\S12_QP_Final-Exam_Deep Learning(702AI0C008)_Semester V_2025-2026.pdf", "answerHtmlSha256": "bcce37de68b300225ac4e9b51c77e21bf85c410f43e6d254b586c593ef499724", "status": "complete", "scope": "current syllabus", "retainedMarks": 5, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q1d", "units": [ 5 ], "type": "case", "sourcePages": [ 1 ], "topic": "RNN context and variable-length translation", "id": "dl-F25-Q1d", "question": "Question (5 marks). Describe sequence processing, context and variable-length sentences in RNNs. Explain their use in language translation, as illustrated by the historical reference to Google Translate.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 5, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 5 ], "answerRoute": "/#tee/question/tee-c11-s3", "answer_route": "/#tee/question/tee-c11-s3", "sourcePath": "sources\\S12_QP_Final-Exam_Deep Learning(702AI0C008)_Semester V_2025-2026.pdf", "answerHtmlSha256": "b44c5d9591d6368c82b0c20a7ed86f46d254213f0479b406ebce3d5fc3379dd3", "status": "complete", "scope": "current syllabus", "retainedMarks": 5, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q2a", "units": [ 2 ], "type": "numerical", "sourcePages": [ 1, 2 ], "topic": "Sigmoid network: forward pass and backpropagation", "id": "dl-F25-Q2a", "question": "Question (10 marks). A 2 → 2 → 1 sigmoid network has all four input-to-hidden weights 0.5, both hidden-to-output weights 0.25, hidden and output biases 0, and target 1. Perform forward and backward passes to update the x₁ → h₁ weight, h₁ → output weight and output bias at η = 0.5. The paper does not state the numerical inputs or loss. F25 Q2a connections with explicitly assumed inputs F25 Q2a connections with explicitly assumed inputs Tap the diagram to enlarge.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 2 ], "answerRoute": "/#tee/question/tee-c11-s4", "answer_route": "/#tee/question/tee-c11-s4", "sourcePath": "sources\\S12_QP_Final-Exam_Deep Learning(702AI0C008)_Semester V_2025-2026.pdf", "answerHtmlSha256": "b2c53a8f57394aef199fe9f55081357ad4d3050058e3eb9459be578245a956ee", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "Original omits inputs and loss. Answer explicitly assumes x=[1,1] and half squared loss; the result is conditional on those assumptions." }, { "label": "Q2b", "units": [ 4 ], "type": "theory", "sourcePages": [ 2 ], "topic": "ResNet; Inception versus AlexNet", "id": "dl-F25-Q2b", "question": "Question (10 marks). Explain how ResNet skip connections influence gradient flow, feature reuse and vanishing gradients. Compare InceptionNet and AlexNet learning and feature extraction.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 4 ], "answerRoute": "/#tee/question/tee-c11-s5", "answer_route": "/#tee/question/tee-c11-s5", "sourcePath": "sources\\S12_QP_Final-Exam_Deep Learning(702AI0C008)_Semester V_2025-2026.pdf", "answerHtmlSha256": "8b7fd802827d104c033372c8119845fb5f6d761cb80bc4eddb50e86b14bd01de", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q3a", "units": [ 2 ], "type": "mixed", "sourcePages": [ 2 ], "topic": "Biological neuron and perceptron epoch", "id": "dl-F25-Q3a", "question": "Question (10 marks). Train one epoch of a single perceptron with η = 1, initial weights and bias 0, threshold 0, on [−1,1,−1] → 0, [−1,1,1] → 1, [1,−1,1] → 1, [−1,1,1] → 1. Then map biological-neuron components to artificial-neuron components.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 2 ], "answerRoute": "/#tee/question/tee-c11-s6", "answer_route": "/#tee/question/tee-c11-s6", "sourcePath": "sources\\S12_QP_Final-Exam_Deep Learning(702AI0C008)_Semester V_2025-2026.pdf", "answerHtmlSha256": "c4bfcdad007489f61ad4c03b63350a001026452124ff7d5baac74a18e6a26ccd", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q3b", "units": [ 4 ], "type": "case", "sourcePages": [ 2 ], "topic": "Transfer learning for face attendance", "id": "dl-F25-Q3b", "question": "Question (10 marks). Design face-recognition attendance using a CNN when the supplied dataset is very small. Discuss ways to address the issue and explain transfer learning.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 4 ], "answerRoute": "/#tee/question/tee-c11-s7", "answer_route": "/#tee/question/tee-c11-s7", "sourcePath": "sources\\S12_QP_Final-Exam_Deep Learning(702AI0C008)_Semester V_2025-2026.pdf", "answerHtmlSha256": "01f3eded992fe4d39bc90d9b3c804e5ba161602af918fa00c9a0183df364416b", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q4a", "units": [ 3 ], "type": "mixed", "sourcePages": [ 2, 3 ], "topic": "Batch counts, momentum iterations and hyperparameters", "id": "dl-F25-Q4a", "question": "Question (10 marks). For N = 1,000 samples and batch size 100, count weight updates per epoch for batch GD, SGD and mini-batch GD. For J(θ) = 0.5 θ², η = 0.1, momentum 0.9, v₀ = 0, θ₀ = 1, compute two momentum iterations. List the required momentum-training hyperparameters.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 3 ], "answerRoute": "/#tee/question/tee-c11-s8", "answer_route": "/#tee/question/tee-c11-s8", "sourcePath": "sources\\S12_QP_Final-Exam_Deep Learning(702AI0C008)_Semester V_2025-2026.pdf", "answerHtmlSha256": "5712cef34910ad4d0a7c355a2145678b39e7d8361be2cd9f3a9315112c3af873", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q4b", "units": [ 5 ], "type": "mixed", "sourcePages": [ 3 ], "topic": "Vanishing gradients and LSTM forget contribution", "id": "dl-F25-Q4b", "question": "Question (10 marks). Define vanishing gradients in RNNs and explain why they occur. Use “The dogs barked at the thief and … were appreciated” to discuss long context. Explain LSTM forgetting/retention, then compute the retained contribution for f = [0.1,0.9], Cprevious = [2,1].", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 5 ], "answerRoute": "/#tee/question/tee-c11-s9", "answer_route": "/#tee/question/tee-c11-s9", "sourcePath": "sources\\S12_QP_Final-Exam_Deep Learning(702AI0C008)_Semester V_2025-2026.pdf", "answerHtmlSha256": "fc22149c85400f461d029eb83c4ff15ada83cc073066684d72e7352fd1d209aa", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q5a", "units": [ 5 ], "type": "mixed", "sourcePages": [ 3 ], "topic": "LSTM gates, cell state and hidden state", "id": "dl-F25-Q5a", "question": "Question (10 marks). State the purpose of each LSTM gate. For “Artificial Intelligence has revolutionized every domain. The work force will have to update themselves to remain relevant,” explain a context change. Calculate C and h given i = 0.7, f = 0.4, o = 0.8, candidate = 0.9, Cprevious = 0.5. Why is LSTM more effective than a simple RNN for language modeling?", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 5 ], "answerRoute": "/#tee/question/tee-c11-s10", "answer_route": "/#tee/question/tee-c11-s10", "sourcePath": "sources\\S12_QP_Final-Exam_Deep Learning(702AI0C008)_Semester V_2025-2026.pdf", "answerHtmlSha256": "8bfddf1deae20a5c39574e0c5c98c36389e959ee6af0f36b87bf7ccd92642afc", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q5b", "units": [ 3 ], "type": "theory", "sourcePages": [ 3 ], "topic": "AdaGrad mechanism and comparison with SGD", "id": "dl-F25-Q5b", "question": "Question (10 marks). Explain AdaGrad, give its update rule and discuss advantages and disadvantages compared with standard SGD.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 3 ], "answerRoute": "/#tee/question/tee-c11-s11", "answer_route": "/#tee/question/tee-c11-s11", "sourcePath": "sources\\S12_QP_Final-Exam_Deep Learning(702AI0C008)_Semester V_2025-2026.pdf", "answerHtmlSha256": "6431ef9229c4c1b925897f1853c7ad0ac1ad469a95ef1a85b88327ff68ff175b", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q6a", "units": [ 4 ], "type": "numerical", "sourcePages": [ 3, 4 ], "topic": "CNN sizes and parameters; stride-2 convolution", "id": "dl-F25-Q6a", "question": "Question (10 marks). Calculate every activation shape and trainable-parameter count for: Input 128 × 128 × 3. → Conv K = 3, filters = 16, stride = 1, padding = 0. → Pool K = 2, stride = 2, padding = 0. → Conv K = 3, filters = 32, stride = 1, padding = 0. → Pool K = 2, stride = 2, padding = 0. → Flatten. → Dense 240 units. → Dense 120 units. → Dense 5 units. Also convolve the input A with the kernel K shown below, valid padding and stride 2. Given matrices Input A Input A 3 1 2 2 1 2 1 2 1 0 3 2 1 0 1 0 3 0 1 2 1 1 0 2 3 Kernel K Kernel K 1 0 −1 2 0 −2 1 0 −1 Stride 2; P = 0 (no padding).", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 4 ], "answerRoute": "/#tee/question/tee-c11-s12", "answer_route": "/#tee/question/tee-c11-s12", "sourcePath": "sources\\S12_QP_Final-Exam_Deep Learning(702AI0C008)_Semester V_2025-2026.pdf", "answerHtmlSha256": "0ccc061784af2d41637206b5e8bbb1aa693a0de3927bba46d0ffd9cb0d5d5774", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q6b", "units": [ 2 ], "type": "theory", "sourcePages": [ 4 ], "topic": "Activation functions and their roles", "id": "dl-F25-Q6b", "question": "Question (10 marks). Explain activations and the role of nonlinearity. Compare sigmoid, tanh, ReLU and Leaky ReLU including advantages and disadvantages.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 2 ], "answerRoute": "/#tee/question/tee-c11-s13", "answer_route": "/#tee/question/tee-c11-s13", "sourcePath": "sources\\S12_QP_Final-Exam_Deep Learning(702AI0C008)_Semester V_2025-2026.pdf", "answerHtmlSha256": "849bab7b4f0cff7e04f2ce8459ce79b6f0d47d3c0600cb2a6e702ad667e26319", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q7a", "units": [ 2 ], "type": "mixed", "sourcePages": [ 4 ], "topic": "Feedforward network roles and ReLU forward pass", "id": "dl-F25-Q7a", "question": "Question (10 marks). Explain feedforward information flow, weights, biases and activations, then calculate the output. Input x = [1, 0]. Each row below lists the incoming weights of one receiving neuron. W₁ is shown below; b₁ = [0.1, 0.2]; W₂ is shown below; b₂ = [0.1]. Hidden and output activations are ReLU. Given matrices for this question Given matrices for this question Tap the diagram to enlarge.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 2 ], "answerRoute": "/#tee/question/tee-c11-s14", "answer_route": "/#tee/question/tee-c11-s14", "sourcePath": "sources\\S12_QP_Final-Exam_Deep Learning(702AI0C008)_Semester V_2025-2026.pdf", "answerHtmlSha256": "19ef500b1bebdf26d1d8b423daa380e6024c6033963578998533b287f5a117fa", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "Receiving-neuron rows give 0.55; the supplied key uses a transposed convention (0.61). Both interpretations are explained." }, { "label": "Q7b", "units": [ 4 ], "type": "theory", "sourcePages": [ 4 ], "topic": "CNN feature hierarchy and applications", "id": "dl-F25-Q7b", "question": "Question (10 marks). Explain hierarchical CNN feature extraction and the roles of convolution, pooling and fully connected layers. Why prefer CNNs over fully connected ANNs for images? Give medical-image and autonomous-vehicle applications and advantages.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 4 ], "answerRoute": "/#tee/question/tee-c11-s15", "answer_route": "/#tee/question/tee-c11-s15", "sourcePath": "sources\\S12_QP_Final-Exam_Deep Learning(702AI0C008)_Semester V_2025-2026.pdf", "answerHtmlSha256": "605ae0eca405e687e704a9b63b76a6cb9d0dd50c3bdfee3b5121e5dfea79f2ad", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" } ] }, { "id": "dl-R26", "subject": "dl", "date": "2026-02-10", "type": "re-exam", "code": "R26", "sourceFile": "S09_QP_Re-Exam_Deep Learning(702AI0C008)_Semester V_2025-2026.pdf", "sourceSha256": "6f31f697f0045bdca36701238c1c35ae5ba3bd42bee352ebc2dc132cb1ad1770", "sourcePages": 4, "visualReview": { "status": "previous verified full-page review reused", "receipt": "audit/dl_nlp_final_20261014.md" }, "pattern": { "durationMinutes": 180, "printedMarks": 140, "attemptMarks": 100, "compulsory": [ "Q1" ], "choose": 4, "from": [ "Q2", "Q3", "Q4", "Q5", "Q6", "Q7" ] }, "questions": [ { "label": "Q1a", "units": [ 1 ], "type": "case", "sourcePages": [ 1 ], "topic": "ML versus DL task choices", "id": "dl-R26-Q1a", "question": "Question (5 marks). Describe two situations favoring DL over traditional ML. Choose ML or DL for house prices from 1,000 numerical records, real-time facial-emotion video recognition, and translation from millions of sentence pairs.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 5, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 1 ], "answerRoute": "/#tee/question/tee-c12-s0", "answer_route": "/#tee/question/tee-c12-s0", "sourcePath": "sources\\S09_QP_Re-Exam_Deep Learning(702AI0C008)_Semester V_2025-2026.pdf", "answerHtmlSha256": "5a1352af493532b82796990f0f4570c9ae585b29c97daf2ea996d0b98ce90835", "status": "complete", "scope": "current syllabus", "retainedMarks": 5, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q1b", "units": [ 2 ], "type": "theory", "sourcePages": [ 1 ], "topic": "Activation functions", "id": "dl-R26-Q1b", "question": "Question (5 marks). Explain how activations affect gradients, convergence and performance. State preferred contexts for sigmoid, tanh, ReLU and Leaky ReLU.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 5, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 2 ], "answerRoute": "/#tee/question/tee-c12-s1", "answer_route": "/#tee/question/tee-c12-s1", "sourcePath": "sources\\S09_QP_Re-Exam_Deep Learning(702AI0C008)_Semester V_2025-2026.pdf", "answerHtmlSha256": "849bab7b4f0cff7e04f2ce8459ce79b6f0d47d3c0600cb2a6e702ad667e26319", "status": "complete", "scope": "current syllabus", "retainedMarks": 5, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q1c", "units": [ 3 ], "type": "mixed", "sourcePages": [ 1 ], "topic": "L1/L2 and specified-rate dropout", "id": "dl-R26-Q1c", "question": "Question (5 marks). Explain L1 and L2 regularization. Apply 50% dropout to outputs [0.8,−0.4,0.6,0.2] using an explicitly assumed mask.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 5, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 3 ], "answerRoute": "/#tee/question/tee-c12-s2", "answer_route": "/#tee/question/tee-c12-s2", "sourcePath": "sources\\S09_QP_Re-Exam_Deep Learning(702AI0C008)_Semester V_2025-2026.pdf", "answerHtmlSha256": "506289987e8ac878171ac9d6e02359a88dc12236bcf6ee2d536728373ae57c96", "status": "complete", "scope": "current syllabus", "retainedMarks": 5, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q1d", "units": [ 5 ], "type": "case", "sourcePages": [ 1 ], "topic": "RNN sentiment classification", "id": "dl-R26-Q1d", "question": "Question (5 marks). Explain an RNN for positive, negative or neutral e-commerce review sentiment. Why are recurrence and word order relevant?", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 5, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 5 ], "answerRoute": "/#tee/question/tee-c12-s3", "answer_route": "/#tee/question/tee-c12-s3", "sourcePath": "sources\\S09_QP_Re-Exam_Deep Learning(702AI0C008)_Semester V_2025-2026.pdf", "answerHtmlSha256": "71c85920812840b6bbe3a2b1e9cbb2760a0572cd3de4015faf47d628294047bf", "status": "complete", "scope": "current syllabus", "retainedMarks": 5, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q2a", "units": [ 2 ], "type": "numerical", "sourcePages": [ 1, 2 ], "topic": "Two-output sigmoid backpropagation", "id": "dl-R26-Q2a", "question": "Question (10 marks). Use the sigmoid network Input x = [0.25, 0.7]. Each row below lists the incoming weights of one receiving neuron. W₁ is shown below; b₁ = [1, 1]; W₂ is shown below; b₂ = [1, 1]. Hidden and output activations are sigmoid. Target y = [0.7, 0.3]. Learning rate η = 1. Here w₁ = x₁ → h₁, w₅ = h₁ → o₁ and w₇ = h₁ → o₂. Run forward and backward passes and update w₁, w₅, w₇. The paper gives no explicit loss formula. Given matrices for this question Given matrices for this question Tap the diagram to enlarge. R26 Q2a weights and both target outputs R26 Q2a weights and both target outputs Tap the diagram to enlarge.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 2 ], "answerRoute": "/#tee/question/tee-c12-s4", "answer_route": "/#tee/question/tee-c12-s4", "sourcePath": "sources\\S09_QP_Re-Exam_Deep Learning(702AI0C008)_Semester V_2025-2026.pdf", "answerHtmlSha256": "64257c8f7ae27216b52d660e9d7a878dba2614f5c2cae32968aaefe2c14afead", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "Loss omitted in original; answer states assumed half summed squared loss." }, { "label": "Q2b", "units": [ 4 ], "type": "case", "sourcePages": [ 2 ], "topic": "AlexNet, Inception and architecture recommendation", "id": "dl-R26-Q2b", "question": "Question (10 marks). Explain AlexNet regularization and training improvements; explain Inception multiscale features and 1 × 1 convolutions; recommend an architecture for large-scale image classification with justification.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 4 ], "answerRoute": "/#tee/question/tee-c12-s5", "answer_route": "/#tee/question/tee-c12-s5", "sourcePath": "sources\\S09_QP_Re-Exam_Deep Learning(702AI0C008)_Semester V_2025-2026.pdf", "answerHtmlSha256": "10b0a7041c92aad9c6b469c66cff04cb30b348bf5eff0768cea1886ac65f943c", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q3a", "units": [ 2 ], "type": "numerical", "sourcePages": [ 2 ], "topic": "Bipolar perceptron epoch", "id": "dl-R26-Q3a", "question": "Question (10 marks). Train a perceptron for one epoch, η = 1, initial weights and bias 0, threshold 0. Samples: [−1,1,−1] → −1; [1,1,1] → 1; [−1,−1,1] → −1; [−1,1,1] → −1. Discuss limitations with examples.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 2 ], "answerRoute": "/#tee/question/tee-c12-s6", "answer_route": "/#tee/question/tee-c12-s6", "sourcePath": "sources\\S09_QP_Re-Exam_Deep Learning(702AI0C008)_Semester V_2025-2026.pdf", "answerHtmlSha256": "2cf015c121e8306dd4afe7786feaf3fe6918b8fabc6e58b140affc02838e0af3", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q3b", "units": [ 4 ], "type": "theory", "sourcePages": [ 2 ], "topic": "Five effects of CNN filter count", "id": "dl-R26-Q3b", "question": "Question (10 marks). How should CNN filter counts be selected? Explain effects on feature extraction, dimensionality, computation/memory, overfitting and hierarchical learning.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 4 ], "answerRoute": "/#tee/question/tee-c12-s7", "answer_route": "/#tee/question/tee-c12-s7", "sourcePath": "sources\\S09_QP_Re-Exam_Deep Learning(702AI0C008)_Semester V_2025-2026.pdf", "answerHtmlSha256": "3b96a951217960f84c4d39c992b4b4563a8f61d23c8afab1acd4539a8707124d", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q4a", "units": [ 3 ], "type": "mixed", "sourcePages": [ 2, 3 ], "topic": "Batch/SGD, learning rate and cubic-objective updates", "id": "dl-R26-Q4a", "question": "Question (10 marks). Compare batch GD and SGD for loss oscillations with a graph, memory and large datasets. Discuss learning rate extremes. For J(θ) = 0.7 θ³, θ₀ = 1, η = 0.1, perform two iterations.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 3 ], "answerRoute": "/#tee/question/tee-c12-s8", "answer_route": "/#tee/question/tee-c12-s8", "sourcePath": "sources\\S09_QP_Re-Exam_Deep Learning(702AI0C008)_Semester V_2025-2026.pdf", "answerHtmlSha256": "02e396ca67f83791dadf4734be7f07d757b86b6e2bb574593b5789d847fb63ed", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q4b", "units": [ 5 ], "type": "theory", "sourcePages": [ 3 ], "topic": "LSTM and GRU; legacy bidirectional-RNN subpart", "id": "dl-R26-Q4b", "question": "Question. Compare LSTM and GRU gates and architecture. Explain the potential computational efficiency of GRU and how its gates can reduce the effect of irrelevant tokens. Syllabus-aligned extract: source R26 Q4b, parts i–iii. Marks for these retained parts are not separately stated.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 5 ], "answerRoute": "/#tee/question/tee-c12-s9", "answer_route": "/#tee/question/tee-c12-s9", "sourcePath": "sources\\S09_QP_Re-Exam_Deep Learning(702AI0C008)_Semester V_2025-2026.pdf", "answerHtmlSha256": "3746941d2812c26aa8c09a3c400051d3b393a1cc3bc89db3a7360ff7bb5d6a01", "status": "current-syllabus extract complete", "scope": "mixed current and legacy", "retainedMarks": null, "excludedTask": "Bidirectional-RNN advantages (iv)", "internalSubparts": [], "notes": "" }, { "label": "Q5a", "units": [ 5 ], "type": "numerical", "sourcePages": [ 3 ], "topic": "RNN hidden-state matrix calculation", "id": "dl-R26-Q5a", "question": "Question (10 marks). Explain feedforward limitations for sequential context and how recurrence helps. Calculate h using tanh for Wxh is shown below, Whh is shown below, x = [0,1], hprevious = [0,1], zero bias. Explain next-word probabilities. Given matrices for this question Given matrices for this question Tap the diagram to enlarge.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 5 ], "answerRoute": "/#tee/question/tee-c12-s10", "answer_route": "/#tee/question/tee-c12-s10", "sourcePath": "sources\\S09_QP_Re-Exam_Deep Learning(702AI0C008)_Semester V_2025-2026.pdf", "answerHtmlSha256": "8285766d0cb299426036e124d391978dbc56905604c6f6dd6899953af8e2ea0d", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q5b", "units": [ 3 ], "type": "theory", "sourcePages": [ 3 ], "topic": "RMSProp mechanism and SGD comparison", "id": "dl-R26-Q5b", "question": "Question (10 marks). Explain RMSProp, its update rule and accumulated squared gradients. Give advantages and disadvantages relative to SGD.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 3 ], "answerRoute": "/#tee/question/tee-c12-s11", "answer_route": "/#tee/question/tee-c12-s11", "sourcePath": "sources\\S09_QP_Re-Exam_Deep Learning(702AI0C008)_Semester V_2025-2026.pdf", "answerHtmlSha256": "462e43910c57774f7f56b02d7b40b75d24170a9f8f55174d76969f14b2c7b5c9", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q6a", "units": [ 4 ], "type": "numerical", "sourcePages": [ 3, 4 ], "topic": "CNN sizes and counts with binary-output ambiguity", "id": "dl-R26-Q6a", "question": "Question (10 marks). For the binary CNN below, calculate output dimensions and trainable parameters after every layer. Input 256 × 256 × 3. → Conv K = 5, filters = 32, stride = 1, padding = 0. → Pool K = 2, stride = 2, padding = 0. → Conv K = 5, filters = 64, stride = 1, padding = 0. → Pool K = 2, stride = 2, padding = 0. → Conv K = 5, filters = 128, stride = 1, padding = 1. → Pool K = 2, stride = 2, padding = 0. → Flatten. → Dense 84 units. → Dense 32 units. → Dense 1 units. The printed output size is unspecified; assume one sigmoid output. Unspecified convolution padding is taken as valid.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 4 ], "answerRoute": "/#tee/question/tee-c12-s12", "answer_route": "/#tee/question/tee-c12-s12", "sourcePath": "sources\\S09_QP_Re-Exam_Deep Learning(702AI0C008)_Semester V_2025-2026.pdf", "answerHtmlSha256": "4182e1cb3cbcef767d80747176902c3b86a5273188b568df053b6012731af25d", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "Binary head unspecified: one sigmoid or two softmax units give different parameter totals; both are stated." }, { "label": "Q6b", "units": [ 2 ], "type": "numerical", "sourcePages": [ 4 ], "topic": "Sigmoid value/derivative and softmax", "id": "dl-R26-Q6b", "question": "Question (10 marks). For a 10-layer network using sigmoid: compute f(6), compute its derivative at 6, explain its effect on backpropagation, and calculate softmax for [2.3,4.5,3.2].", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 2 ], "answerRoute": "/#tee/question/tee-c12-s13", "answer_route": "/#tee/question/tee-c12-s13", "sourcePath": "sources\\S09_QP_Re-Exam_Deep Learning(702AI0C008)_Semester V_2025-2026.pdf", "answerHtmlSha256": "eb97cdaea36da81ea87797506deb50802c01b2bf3811fc0fb54d0f2f92efcf08", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q7a", "units": [ 2 ], "type": "mixed", "sourcePages": [ 4 ], "topic": "Feedforward roles and sigmoid forward pass", "id": "dl-R26-Q7a", "question": "Question (10 marks). Compare single-layer and multilayer feedforward networks and explain hidden nonlinearity. Compute the output for Input x = [0.5, 0.2, 0.1]. Each row below lists the incoming weights of one receiving neuron. W₁ is shown below; b₁ = [0.1, 0.2]; W₂ is shown below; b₂ = [0.5]. Hidden and output activations are sigmoid. Given matrices for this question Given matrices for this question Tap the diagram to enlarge.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 2 ], "answerRoute": "/#tee/question/tee-c12-s14", "answer_route": "/#tee/question/tee-c12-s14", "sourcePath": "sources\\S09_QP_Re-Exam_Deep Learning(702AI0C008)_Semester V_2025-2026.pdf", "answerHtmlSha256": "d853eb03ed1b4e25816cc001e0d8489d8c0c340c41d22139f9f0a826bb588e53", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q7b", "units": [ 4 ], "type": "case", "sourcePages": [ 4 ], "topic": "Transfer learning and its application", "id": "dl-R26-Q7b", "question": "Question (10 marks). Explain transfer learning with CNNs and how to adapt a pretrained model to a new image classification task, including an example model.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 4 ], "answerRoute": "/#tee/question/tee-c12-s15", "answer_route": "/#tee/question/tee-c12-s15", "sourcePath": "sources\\S09_QP_Re-Exam_Deep Learning(702AI0C008)_Semester V_2025-2026.pdf", "answerHtmlSha256": "c8cfaa0fe6dee6750c1092bc6e14f565baeb590a9ef7abc462d3817b793d6109", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" } ] }, { "id": "dl-F23", "subject": "dl", "date": "2023-12-02", "type": "final", "code": "F23", "sourceFile": "S17_Deep_Learning__Final_exam_2023-24_EyPdFZ3lIK.pdf", "sourceSha256": "017671e0097f1bfe6baea738b916dc9d38d3a78006c85831654fd4d4e7109e41", "sourcePages": 5, "visualReview": { "status": "fresh: every original page viewed 2026-10-14", "receipt": "audit/dl_nlp_all_pyq_review_20261014.md" }, "pattern": { "durationMinutes": 180, "printedMarks": 140, "attemptMarks": 100, "compulsory": [ "Q1" ], "choose": 4, "from": [ "Q2", "Q3", "Q4", "Q5", "Q6", "Q7" ] }, "questions": [ { "label": "Q1a", "units": [ 1 ], "type": "theory", "sourcePages": [ 1 ], "topic": "Deep learning features and limitations", "id": "dl-F23-Q1a", "question": "Question (5 marks). State the features and limitations of deep learning.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 5, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 1 ], "answerRoute": "/#tee/question/tee-c13-s0", "answer_route": "/#tee/question/tee-c13-s0", "sourcePath": "sources\\S17_Deep_Learning__Final_exam_2023-24_EyPdFZ3lIK.pdf", "answerHtmlSha256": "6bcfe1e1e0962baeff856d4a81343fd86a7b38da77268d38a404cccf30de4e19", "status": "complete", "scope": "current syllabus", "retainedMarks": 5, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q1b", "units": [ 2, 3 ], "type": "theory", "sourcePages": [ 1 ], "topic": "ANN hyperparameters and significance", "id": "dl-F23-Q1b", "question": "Question (5 marks). Identify the hyperparameters to tune when training an ANN and explain their significance.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 5, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 2, 3 ], "answerRoute": "/#tee/question/tee-c13-s1", "answer_route": "/#tee/question/tee-c13-s1", "sourcePath": "sources\\S17_Deep_Learning__Final_exam_2023-24_EyPdFZ3lIK.pdf", "answerHtmlSha256": "d2431e3b21549f686c4bd30f85bb3625d3cff8f495357f809e7717e7d932a312", "status": "complete", "scope": "current syllabus", "retainedMarks": 5, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q1c", "units": [ 5 ], "type": "theory", "sourcePages": [ 1 ], "topic": "Unrolled RNN and BPTT", "id": "dl-F23-Q1c", "question": "Question (5 marks). Draw a time-unrolled RNN and briefly explain training by backpropagation through time.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 5, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 5 ], "answerRoute": "/#tee/question/tee-c13-s2", "answer_route": "/#tee/question/tee-c13-s2", "sourcePath": "sources\\S17_Deep_Learning__Final_exam_2023-24_EyPdFZ3lIK.pdf", "answerHtmlSha256": "0ad1400a603cca1c108b1226e63192ff381d8dfda2355540f958c7fa2686ee35", "status": "complete", "scope": "current syllabus", "retainedMarks": 5, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q1d", "units": [ 4 ], "type": "numerical", "sourcePages": [ 1 ], "topic": "Convolution map", "id": "dl-F23-Q1d", "question": "Question (5 marks). Compute the convolution using the input A and kernel K shown below using padding 0 and stride 1. Given matrices Input A Input A 10 9 5 6 5 7 6 7 3 3 2 8 3 2 6 7 Kernel K Kernel K 1 1 1 1 2 1 1 1 1 Stride 1; P = 0 (no padding).", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 5, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 4 ], "answerRoute": "/#tee/question/tee-c13-s3", "answer_route": "/#tee/question/tee-c13-s3", "sourcePath": "sources\\S17_Deep_Learning__Final_exam_2023-24_EyPdFZ3lIK.pdf", "answerHtmlSha256": "438f659fb7df01082d6999a9174b660123cbd52b692630c9be37054fa29e4fe2", "status": "complete", "scope": "current syllabus", "retainedMarks": 5, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q2a", "units": [ 2 ], "type": "numerical", "sourcePages": [ 2 ], "topic": "Binary-input step-activation network", "id": "dl-F23-Q2a", "question": "Question (10 marks). A feedforward network has inputs 1,2; hidden nodes 3,4; outputs 5,6. Weights: w13 = −2, w23 = 3, w14 = 4, w24 = −1, w35 = 1, w45 = −1, w36 = −1, w46 = 1; all biases 0. All nodes 3–6 use step(z) = 1 if z≥0 else 0. Find(y5, y6) for inputs(0,0), (1,0), (0,1), (1,1).", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 2 ], "answerRoute": "/#tee/question/tee-c13-s4", "answer_route": "/#tee/question/tee-c13-s4", "sourcePath": "sources\\S17_Deep_Learning__Final_exam_2023-24_EyPdFZ3lIK.pdf", "answerHtmlSha256": "c1a9822527d36ebac5a2658a508a78d883768c8c70ab7efa5eba8492d5160c6e", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q2b", "units": [ 2, 1 ], "type": "mixed", "sourcePages": [ 3 ], "topic": "Activation necessity; ML/DL facial-recognition choice", "id": "dl-F23-Q2b", "question": "Question (10 marks). (i) Justify the need for activations and explain any two. (ii) Choose ML or DL for facial recognition using 227 × 227 × 1 images and state the advantages.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 2, 1 ], "answerRoute": "/#tee/question/tee-c13-s5", "answer_route": "/#tee/question/tee-c13-s5", "sourcePath": "sources\\S17_Deep_Learning__Final_exam_2023-24_EyPdFZ3lIK.pdf", "answerHtmlSha256": "c2d469af7dcbab2462f9f5b2227731a279f796c06611a5ef7001c5dfa9081369", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [ { "label": "i", "marks": 5, "topic": "Activation necessity and any two activations" }, { "label": "ii", "marks": 5, "topic": "Facial-recognition model choice and advantages" } ], "notes": "" }, { "label": "Q3a", "units": [ 2 ], "type": "numerical", "sourcePages": [ 3 ], "topic": "Sigmoid output-weight backpropagation", "id": "dl-F23-Q3a", "question": "Question (10 marks). For x = [1,0,1], target 1, η = 0.9, use sigmoid at hidden and output layers. Weights w14 = .2, w15 = −.3, w24 = .4, w25 = .1, w34 = −.5, w35 = .2, w46 = −.3, w56 = −.2, biases at 4,5,6 are−.4,.2,.1. Update w46 and w56 using backpropagation. F23 Q3a connections with the supplied biases F23 Q3a connections with the supplied biases Tap the diagram to enlarge.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 2 ], "answerRoute": "/#tee/question/tee-c13-s6", "answer_route": "/#tee/question/tee-c13-s6", "sourcePath": "sources\\S17_Deep_Learning__Final_exam_2023-24_EyPdFZ3lIK.pdf", "answerHtmlSha256": "c2b720ae0cec312ef6a9c86b4e391a61a7e016ab8a2226f971e69e906c3caef2", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q3b", "units": [ 3 ], "type": "theory", "sourcePages": [ 3 ], "topic": "Adam advantages versus SGD", "id": "dl-F23-Q3b", "question": "Question (10 marks). Explain key advantages of Adam for ANN training and how it differs from SGD.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 3 ], "answerRoute": "/#tee/question/tee-c13-s7", "answer_route": "/#tee/question/tee-c13-s7", "sourcePath": "sources\\S17_Deep_Learning__Final_exam_2023-24_EyPdFZ3lIK.pdf", "answerHtmlSha256": "cc47a04a3606c6de2a28ad8d93451c396d9f31c8fccd21da8507676b309493cf", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q4a", "units": [ 4 ], "type": "numerical", "sourcePages": [ 4 ], "topic": "AlexNet-style layer dimensions and parameters", "id": "dl-F23-Q4a", "question": "Question (10 marks). Determine feature-map dimensions and parameters at every layer of the pictured AlexNet-style CNN. The normalized architecture is Input 227 × 227 × 3. → Conv K = 11, filters = 96, stride = 4, padding = 0. → Pool K = 3, stride = 2, padding = 0. → Conv K = 5, filters = 256, stride = 1, padding = 2. → Pool K = 3, stride = 2, padding = 0. → Conv K = 3, filters = 384, stride = 1, padding = 1. → Conv K = 3, filters = 384, stride = 1, padding = 1. → Conv K = 3, filters = 256, stride = 1, padding = 1. → Pool K = 3, stride = 2, padding = 0. → Flatten. → Dense 4096 units. → Dense 4096 units. → Dense 1000 units.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 4 ], "answerRoute": "/#tee/question/tee-c13-s8", "answer_route": "/#tee/question/tee-c13-s8", "sourcePath": "sources\\S17_Deep_Learning__Final_exam_2023-24_EyPdFZ3lIK.pdf", "answerHtmlSha256": "8d3b33e50f309150bc38a9009a6e496893b54bc0fbda8b7136577ec342a4d92b", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q4b", "units": [ 5 ], "type": "case", "sourcePages": [ 4 ], "topic": "LSTM autocomplete architecture and token probabilities", "id": "dl-F23-Q4b", "question": "Question (10 marks). Design an LSTM autocomplete model for “Maharashtra is a state in the western part of India. The capital of Maharashtra is ___.” Draw the model and write conditional probabilities at each time step.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 5 ], "answerRoute": "/#tee/question/tee-c13-s9", "answer_route": "/#tee/question/tee-c13-s9", "sourcePath": "sources\\S17_Deep_Learning__Final_exam_2023-24_EyPdFZ3lIK.pdf", "answerHtmlSha256": "6aeaa46ead8395d1516c5e98b40812c91037c3338ad3c3482ebdd46d45f1c428", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q5a", "units": [ 4 ], "type": "theory", "sourcePages": [ 4 ], "topic": "Inception branches and 1x1 convolution", "id": "dl-F23-Q5a", "question": "Question (10 marks). For Inception V 1, justify different kernel sizes in one block, explain 1 × 1 convolutions and sketch one block.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 4 ], "answerRoute": "/#tee/question/tee-c13-s10", "answer_route": "/#tee/question/tee-c13-s10", "sourcePath": "sources\\S17_Deep_Learning__Final_exam_2023-24_EyPdFZ3lIK.pdf", "answerHtmlSha256": "bb45050abc5cba87ba44dc6bddf82c891e442b27e16ae5ff9101801b2b4cfd07", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q5b", "units": [ 2 ], "type": "mixed", "sourcePages": [ 4 ], "topic": "Linear separability; XOR and XNOR construction", "id": "dl-F23-Q5b", "question": "Question (10 marks). Explain why a single perceptron cannot classify nonlinearly separable data. Implement XOR and XNOR using suitable perceptrons.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 2 ], "answerRoute": "/#tee/question/tee-c13-s11", "answer_route": "/#tee/question/tee-c13-s11", "sourcePath": "sources\\S17_Deep_Learning__Final_exam_2023-24_EyPdFZ3lIK.pdf", "answerHtmlSha256": "ddc1722e5ab759fafcd13b9bdc6f7624e32aa85b8faa50b0dd4ba0bd6e9db0db", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [ { "label": "i", "marks": 5, "topic": "Nonlinear separability limitation" }, { "label": "ii", "marks": 5, "topic": "XOR and XNOR implementation" } ], "notes": "" }, { "label": "Q6a", "units": [ 5 ], "type": "theory", "sourcePages": [ 4, 5 ], "topic": "LSTM gates; source mixes reset/update terminology", "id": "dl-F23-Q6a", "question": "Question (10 marks). For the pictured LSTM time step, explain the forget, “update” and output gates, develop “update and reset” equations, and discuss long sequence lengths. The source mixes LSTM and GRU terminology.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 5 ], "answerRoute": "/#tee/question/tee-c13-s12", "answer_route": "/#tee/question/tee-c13-s12", "sourcePath": "sources\\S17_Deep_Learning__Final_exam_2023-24_EyPdFZ3lIK.pdf", "answerHtmlSha256": "1bf3fd1808dfda8726cc79e143dab700a99d20541eb3981772c7b2038df80d44", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "Printed LSTM diagram uses update-gate wording and asks reset equations. Answer separates LSTM and GRU rather than inventing an LSTM reset gate." }, { "label": "Q6b", "units": [ 4 ], "type": "theory", "sourcePages": [ 5 ], "topic": "ResNet gradient flow and block outputs", "id": "dl-F23-Q6b", "question": "Question (10 marks). Explain how ResNet addresses vanishing gradients. Compare a block output with and without a residual connection.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 4 ], "answerRoute": "/#tee/question/tee-c13-s13", "answer_route": "/#tee/question/tee-c13-s13", "sourcePath": "sources\\S17_Deep_Learning__Final_exam_2023-24_EyPdFZ3lIK.pdf", "answerHtmlSha256": "f00aab44bb8cbda086f809b6981de09a4e7f105c105b0e1f27d32836bbfc33c8", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q7a", "units": [ 5 ], "type": "case", "sourcePages": [ 5 ], "topic": "Paragraph sentiment model", "id": "dl-F23-Q7a", "question": "Question (10 marks). Design and sketch a sentiment-analysis model whose input is a small paragraph. Explain how its output is generated.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 5 ], "answerRoute": "/#tee/question/tee-c13-s14", "answer_route": "/#tee/question/tee-c13-s14", "sourcePath": "sources\\S17_Deep_Learning__Final_exam_2023-24_EyPdFZ3lIK.pdf", "answerHtmlSha256": "6499885c98d2b9b8e207291cd363a7175fd7eac4e52b4893ec72cbfee01f818d", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q7b", "units": [ 3 ], "type": "theory", "sourcePages": [ 5 ], "topic": "Bias-variance tradeoff and intervention table", "id": "dl-F23-Q7b", "question": "Question (10 marks). Explain the bias-variance tradeoff. Fill effects on bias and variance for weight regularization, more hidden units, dropout and more training data from the same distribution.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 3 ], "answerRoute": "/#tee/question/tee-c13-s15", "answer_route": "/#tee/question/tee-c13-s15", "sourcePath": "sources\\S17_Deep_Learning__Final_exam_2023-24_EyPdFZ3lIK.pdf", "answerHtmlSha256": "c41cb97fd97c5349d55be15244a39712636fc5f3aada59bd4392d91c66282283", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [ { "label": "i", "marks": 5, "topic": "Bias-variance tradeoff" }, { "label": "ii", "marks": 5, "topic": "Four intervention rows" } ], "notes": "" } ] }, { "id": "dl-R24", "subject": "dl", "date": "2024-02-13", "type": "re-exam", "code": "R24", "sourceFile": "S14_Deep_Learning___Re-exam_2023-24_oxnB58Pwaa.pdf", "sourceSha256": "3df2c1997c7790a13ab181eae53997097bc628f2c0949eaf6810f8324fa07f23", "sourcePages": 5, "visualReview": { "status": "fresh: every original page viewed 2026-10-14", "receipt": "audit/dl_nlp_all_pyq_review_20261014.md" }, "pattern": { "durationMinutes": 180, "printedMarks": 140, "attemptMarks": 100, "compulsory": [ "Q1" ], "choose": 4, "from": [ "Q2", "Q3", "Q4", "Q5", "Q6", "Q7" ] }, "questions": [ { "label": "Q1a", "units": [ 1 ], "type": "theory", "sourcePages": [ 1 ], "topic": "Drivers of deep learning adoption", "id": "dl-R24-Q1a", "question": "Question (5 marks). What factors drive widespread adoption of deep learning?", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 5, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 1 ], "answerRoute": "/#tee/question/tee-c14-s0", "answer_route": "/#tee/question/tee-c14-s0", "sourcePath": "sources\\S14_Deep_Learning___Re-exam_2023-24_oxnB58Pwaa.pdf", "answerHtmlSha256": "6bcfe1e1e0962baeff856d4a81343fd86a7b38da77268d38a404cccf30de4e19", "status": "complete", "scope": "current syllabus", "retainedMarks": 5, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q1b", "units": [ 2 ], "type": "numerical", "sourcePages": [ 1 ], "topic": "10-5-5-5-4 dense network parameter count", "id": "dl-R24-Q1b", "question": "Question (5 marks). Draw an animal-classification network with 10 inputs, three hidden layers of 5 neurons each and 4 output classes. Every hidden/output neuron has a bias. Count all trainable parameters.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 5, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 2 ], "answerRoute": "/#tee/question/tee-c14-s1", "answer_route": "/#tee/question/tee-c14-s1", "sourcePath": "sources\\S14_Deep_Learning___Re-exam_2023-24_oxnB58Pwaa.pdf", "answerHtmlSha256": "828574cd2c78550bb6f1fc7fa65adea6a1213706758a724092880e414d15326a", "status": "complete", "scope": "current syllabus", "retainedMarks": 5, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q1c", "units": [ 3 ], "type": "case", "sourcePages": [ 1 ], "topic": "Overfitting diagnosis and remedies", "id": "dl-R24-Q1c", "question": "Question (5 marks). A deep network strongly overfits its training set and performs poorly on unseen data. Explain remedies.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 5, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 3 ], "answerRoute": "/#tee/question/tee-c14-s2", "answer_route": "/#tee/question/tee-c14-s2", "sourcePath": "sources\\S14_Deep_Learning___Re-exam_2023-24_oxnB58Pwaa.pdf", "answerHtmlSha256": "4bde4e3fa8d0eba055e0d44b54f1b4719e7a0cdce9ddb62b4dbb83514afb368e", "status": "complete", "scope": "current syllabus", "retainedMarks": 5, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q1d", "units": [ 4 ], "type": "numerical", "sourcePages": [ 2 ], "topic": "Max and average pooling", "id": "dl-R24-Q1d", "question": "Question (5 marks). For feature map Matrix 1 shown below, find max and average pooling with 2 × 2 windows, stride 2. Given matrices for this question Given matrices for this question Tap the diagram to enlarge.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 5, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 4 ], "answerRoute": "/#tee/question/tee-c14-s3", "answer_route": "/#tee/question/tee-c14-s3", "sourcePath": "sources\\S14_Deep_Learning___Re-exam_2023-24_oxnB58Pwaa.pdf", "answerHtmlSha256": "1c00052534bc6852cfa7aa4a171ad23a7cf855b3e2e5011cbbe076bbf77202c2", "status": "complete", "scope": "current syllabus", "retainedMarks": 5, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q2a", "units": [ 2 ], "type": "numerical", "sourcePages": [ 2, 3 ], "topic": "Four-hidden-neuron sigmoid MSE and update", "id": "dl-R24-Q2a", "question": "Question (10 marks). A 4 → 4 → 1 sigmoid network has input [.4,.3,.5,−.4]. The printed W₁ is shown below, W₂ = [.5,.7,−.6,.2]; hidden bias 0.6, output bias 0.2, target 0.9, η = .5. Compute mean squared error and update the connection h₁ → output. In the graph θ ij connects input i to hidden j. Given matrices for this question Given matrices for this question Tap the diagram to enlarge. R24 Q2a receiving-neuron rows after transposing the source convention R24 Q2a receiving-neuron rows after transposing the source convention Tap the diagram to enlarge.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 2 ], "answerRoute": "/#tee/question/tee-c14-s4", "answer_route": "/#tee/question/tee-c14-s4", "sourcePath": "sources\\S14_Deep_Learning___Re-exam_2023-24_oxnB58Pwaa.pdf", "answerHtmlSha256": "870a183a00f384104af23b3d996f869426ca20e340d4e868ed8642990fea8bd9", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q2b", "units": [ 3 ], "type": "theory", "sourcePages": [ 3 ], "topic": "Momentum/Nesterov; high momentum; excluded convex rates", "id": "dl-R24-Q2b", "question": "Question. Compare standard momentum and Nesterov accelerated gradient using their gradient-evaluation points and update equations. Explain what may happen when momentum is extremely close to 1, for example 0.9999. Syllabus-aligned extract: source R24 Q2b. The current-syllabus optimizer comparison and high-momentum discussion are retained; marks for this extract are not separately stated.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 3 ], "answerRoute": "/#tee/question/tee-c14-s5", "answer_route": "/#tee/question/tee-c14-s5", "sourcePath": "sources\\S14_Deep_Learning___Re-exam_2023-24_oxnB58Pwaa.pdf", "answerHtmlSha256": "bc4a2b98ca27c8454ecae33503c2d694516286ee97eac4cb88aa83d42d3715eb", "status": "current-syllabus extract complete", "scope": "mixed current and legacy", "retainedMarks": null, "excludedTask": "Asymptotic convergence-rate comparison on convex problems (part of i)", "internalSubparts": [ { "label": "i", "marks": 5, "topic": "Momentum/Nesterov comparison plus legacy convex convergence rate" }, { "label": "ii", "marks": 5, "topic": "Momentum extremely close to one" } ], "notes": "" }, { "label": "Q3a", "units": [ 2 ], "type": "mixed", "sourcePages": [ 3 ], "topic": "Activation comparisons and softmax", "id": "dl-R24-Q3a", "question": "Question (10 marks). Differentiate sigmoid and ReLU(4). Explain why sigmoid/tanh are often avoided in deep hidden layers(3). Compute softmax of [2.1,4.8,3.5](3).", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 2 ], "answerRoute": "/#tee/question/tee-c14-s6", "answer_route": "/#tee/question/tee-c14-s6", "sourcePath": "sources\\S14_Deep_Learning___Re-exam_2023-24_oxnB58Pwaa.pdf", "answerHtmlSha256": "71d2ce021d8a1c923aa1cbc442c2b7cf51cb7a6b9b04c1f9250da9ce6526465e", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [ { "label": "i", "marks": 4, "topic": "Sigmoid versus ReLU" }, { "label": "ii", "marks": 3, "topic": "Hidden-layer sigmoid/tanh limitation" }, { "label": "iii", "marks": 3, "topic": "Softmax calculation" } ], "notes": "" }, { "label": "Q3b", "units": [ 3 ], "type": "theory", "sourcePages": [ 3 ], "topic": "Regularization importance and techniques", "id": "dl-R24-Q3b", "question": "Question (10 marks). Explain why regularization matters, list techniques and elaborate one.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 3 ], "answerRoute": "/#tee/question/tee-c14-s7", "answer_route": "/#tee/question/tee-c14-s7", "sourcePath": "sources\\S14_Deep_Learning___Re-exam_2023-24_oxnB58Pwaa.pdf", "answerHtmlSha256": "4b6564d54e1ae6ef4bc5a99098a7ba3f9e0b7be76b3afc85aa7f9e4d22c12bf2", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q4a", "units": [ 2 ], "type": "numerical", "sourcePages": [ 3 ], "topic": "Perceptron training for two epochs", "id": "dl-R24-Q4a", "question": "Question (10 marks). Train a perceptron for 2 epochs with η = 1, initial weights and bias 0, threshold 0. Samples in order: [1,1,1,1] → 1; [−1,1,−1,−1] → 1; [1,1,1,−1] → −1; [1,−1,−1,1] → −1. Bias input is 1.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 2 ], "answerRoute": "/#tee/question/tee-c14-s8", "answer_route": "/#tee/question/tee-c14-s8", "sourcePath": "sources\\S14_Deep_Learning___Re-exam_2023-24_oxnB58Pwaa.pdf", "answerHtmlSha256": "911733109c7807d16e3245134415b6b50993eca819545b560a5431aeb82d62d3", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q4b", "units": [ 4 ], "type": "numerical", "sourcePages": [ 4 ], "topic": "CNN sizes, weights and biases", "id": "dl-R24-Q4b", "question": "Question (10 marks). Calculate weights, biases and feature-map sizes of the CNN: Input 227 × 227 × 3. → Conv K = 9, filters = 32, stride = 1, padding = 0. → Pool K = 2, stride = 2, padding = 0. → Conv K = 5, filters = 64, stride = 1, padding = 0. → Pool K = 2, stride = 2, padding = 0. → Conv K = 5, filters = 64, stride = 1, padding = 0. → Pool K = 2, stride = 2, padding = 0. → Flatten. → Dense 120 units. → Dense 84 units. → Dense 10 units.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 4 ], "answerRoute": "/#tee/question/tee-c14-s9", "answer_route": "/#tee/question/tee-c14-s9", "sourcePath": "sources\\S14_Deep_Learning___Re-exam_2023-24_oxnB58Pwaa.pdf", "answerHtmlSha256": "97bcfe441a05d8350aeaccb75535d8ede2ea5170ce813b0858b2714a0a384f5b", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q5a", "units": [ 4 ], "type": "case", "sourcePages": [ 4 ], "topic": "Replacing architecture to improve gradient flow", "id": "dl-R24-Q5a", "question": "Question (10 marks). A GoogLeNet classifier trains slowly and vanishing gradients are suspected. Suggest an architecture change and explain its important features.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 4 ], "answerRoute": "/#tee/question/tee-c14-s10", "answer_route": "/#tee/question/tee-c14-s10", "sourcePath": "sources\\S14_Deep_Learning___Re-exam_2023-24_oxnB58Pwaa.pdf", "answerHtmlSha256": "05b333717fc57edcc1ec9c024ab10109503932d97b3f7f46fc6e5a096edacd98", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q5b", "units": [ 1, 5 ], "type": "case", "sourcePages": [ 4 ], "topic": "X-ray DL justification; RNN pros and cons", "id": "dl-R24-Q5b", "question": "Question (10 marks). (i) Justify DL for chest-cancer diagnosis from X-ray images(5). (ii) List RNN advantages and disadvantages(5).", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 1, 5 ], "answerRoute": "/#tee/question/tee-c14-s11", "answer_route": "/#tee/question/tee-c14-s11", "sourcePath": "sources\\S14_Deep_Learning___Re-exam_2023-24_oxnB58Pwaa.pdf", "answerHtmlSha256": "f6f2a53ae7ee332bdb877f99341c555ca9aba7cd2d806d2834c5112c96237452", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [ { "label": "i", "marks": 5, "topic": "Chest-X-ray DL justification" }, { "label": "ii", "marks": 5, "topic": "RNN advantages and disadvantages" } ], "notes": "" }, { "label": "Q6a", "units": [ 5 ], "type": "theory", "sourcePages": [ 4, 5 ], "topic": "GRU update and reset gates", "id": "dl-R24-Q6a", "question": "Question (10 marks). For a GRU time step, explain the significance of update and reset gates and develop their equations.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 5 ], "answerRoute": "/#tee/question/tee-c14-s12", "answer_route": "/#tee/question/tee-c14-s12", "sourcePath": "sources\\S14_Deep_Learning___Re-exam_2023-24_oxnB58Pwaa.pdf", "answerHtmlSha256": "3746941d2812c26aa8c09a3c400051d3b393a1cc3bc89db3a7360ff7bb5d6a01", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q6b", "units": [ 4 ], "type": "case", "sourcePages": [ 5 ], "topic": "Pretrained vegetable classifier", "id": "dl-R24-Q6b", "question": "Question (10 marks). A supermarket vegetable classifier has 10 classes but few examples per class and limited training resources. Explain different ways to use a pretrained model.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 4 ], "answerRoute": "/#tee/question/tee-c14-s13", "answer_route": "/#tee/question/tee-c14-s13", "sourcePath": "sources\\S14_Deep_Learning___Re-exam_2023-24_oxnB58Pwaa.pdf", "answerHtmlSha256": "364eac7294fca33553d08c533a3b795e79135404d82ddd0e0c488d8908de3a2c", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q7a", "units": [ 5 ], "type": "theory", "sourcePages": [ 5 ], "topic": "LSTM architecture", "id": "dl-R24-Q7a", "question": "Question (10 marks). Draw and explain LSTM architecture.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 5 ], "answerRoute": "/#tee/question/tee-c14-s14", "answer_route": "/#tee/question/tee-c14-s14", "sourcePath": "sources\\S14_Deep_Learning___Re-exam_2023-24_oxnB58Pwaa.pdf", "answerHtmlSha256": "ffc4be82157b9ea79fc6ab584ef0c5a034bb5441e23912fc8573fd547fbe0d51", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q7b", "units": [ 5 ], "type": "theory", "sourcePages": [ 5 ], "topic": "RNN input/output sequence architectures", "id": "dl-R24-Q7b", "question": "Question (10 marks). Explain different RNN input/output sequence types with architecture sketches.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 5 ], "answerRoute": "/#tee/question/tee-c14-s15", "answer_route": "/#tee/question/tee-c14-s15", "sourcePath": "sources\\S14_Deep_Learning___Re-exam_2023-24_oxnB58Pwaa.pdf", "answerHtmlSha256": "03dc36352b4c18c6e66cd7d0d5d14ce290525af6dd53e5056856d865ecda4815", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" } ] }, { "id": "dl-F22", "subject": "dl", "date": "2022-11-19", "type": "final", "code": "F22", "sourceFile": "S16_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Final_Exam_rTqlPdZM6a.pdf", "sourceSha256": "39d3f520183f42336d1a7671aa946f3a6555484b1c5f6d7b8cdcb513efbda374", "sourcePages": 5, "visualReview": { "status": "fresh: every original page viewed 2026-10-14", "receipt": "audit/dl_nlp_all_pyq_review_20261014.md" }, "pattern": { "durationMinutes": 180, "printedMarks": 140, "attemptMarks": 100, "compulsory": [ "Q1" ], "choose": 4, "from": [ "Q2", "Q3", "Q4", "Q5", "Q6", "Q7" ] }, "questions": [ { "label": "Q1a", "units": [ 4 ], "type": "numerical", "sourcePages": [ 1 ], "topic": "Stride-2 convolution and detected feature", "id": "dl-F22-Q1a", "question": "Question (4 marks). A normalized 5 × 5 image has every row [0.5,0.5,0.3,0.3,0.3]. Convolve it with a 3 × 3 filter whose every row is [1,0,−1], stride 2, no padding. Compute the output and identify the feature.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 4, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 4 ], "answerRoute": "/#tee/question/tee-c15-s0", "answer_route": "/#tee/question/tee-c15-s0", "sourcePath": "sources\\S16_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Final_Exam_rTqlPdZM6a.pdf", "answerHtmlSha256": "02fc53253c74028972d330514432b52a29a6ffee4abc4d23bd5af5289d5f5871", "status": "complete", "scope": "current syllabus", "retainedMarks": 4, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q1b", "units": [ 2 ], "type": "numerical", "sourcePages": [ 1 ], "topic": "Signed gradient-descent update", "id": "dl-F22-Q1b", "question": "Question (4 marks). Weights are 0.55 and 0.65; their gradients are+0.3 and−0.25. With η = 0.1, compute gradient-descent changes and new weights.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 4, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 2 ], "answerRoute": "/#tee/question/tee-c15-s1", "answer_route": "/#tee/question/tee-c15-s1", "sourcePath": "sources\\S16_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Final_Exam_rTqlPdZM6a.pdf", "answerHtmlSha256": "b8ad8a0ec4e6bdf5718ad1b17919b4fc1f41aa4c08ee0741eb0358e788d5a0e6", "status": "complete", "scope": "current syllabus", "retainedMarks": 4, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q1c", "units": [ 1 ], "type": "theory", "sourcePages": [ 2 ], "topic": "ML and DL comparison", "id": "dl-F22-Q1c", "question": "Question (4 marks). Compare ML and DL algorithms.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 4, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 1 ], "answerRoute": "/#tee/question/tee-c15-s2", "answer_route": "/#tee/question/tee-c15-s2", "sourcePath": "sources\\S16_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Final_Exam_rTqlPdZM6a.pdf", "answerHtmlSha256": "6bcfe1e1e0962baeff856d4a81343fd86a7b38da77268d38a404cccf30de4e19", "status": "complete", "scope": "current syllabus", "retainedMarks": 4, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q1d", "units": [ 2 ], "type": "theory", "sourcePages": [ 2 ], "topic": "Sigmoid versus tanh", "id": "dl-F22-Q1d", "question": "Question (4 marks). Compare sigmoid and tanh, emphasizing tanh’s advantage.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 4, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 2 ], "answerRoute": "/#tee/question/tee-c15-s3", "answer_route": "/#tee/question/tee-c15-s3", "sourcePath": "sources\\S16_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Final_Exam_rTqlPdZM6a.pdf", "answerHtmlSha256": "b4d6c5fa71a2e263be53e7bbd4f7f10a8debff0bc0a4dcd4109278367390a45f", "status": "complete", "scope": "current syllabus", "retainedMarks": 4, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q1e", "units": [ 5 ], "type": "case", "sourcePages": [ 2 ], "topic": "Image-captioning RNN architecture", "id": "dl-F22-Q1e", "question": "Question (4 marks). Suggest and sketch an RNN architecture for image captioning.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 4, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 5 ], "answerRoute": "/#tee/question/tee-c15-s4", "answer_route": "/#tee/question/tee-c15-s4", "sourcePath": "sources\\S16_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Final_Exam_rTqlPdZM6a.pdf", "answerHtmlSha256": "bc05a89416ac27dc166ee4de9b6e125e78de1e1ac81997cd09202e056d65005e", "status": "complete", "scope": "current syllabus", "retainedMarks": 4, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q2a", "units": [ 5 ], "type": "case", "sourcePages": [ 2 ], "topic": "GRU advantages and subject memory in a sentence", "id": "dl-F22-Q2a", "question": "Question (10 marks). Explain GRU advantages over a normal RNN. For “The student who did not submit the assignment was bound to loose marks,” show how memory and update gates can retain the subject before predicting “was”.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 5 ], "answerRoute": "/#tee/question/tee-c15-s5", "answer_route": "/#tee/question/tee-c15-s5", "sourcePath": "sources\\S16_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Final_Exam_rTqlPdZM6a.pdf", "answerHtmlSha256": "4d62f320f5a83c08660e443325f137507455581ab19ac6e30da14eded5bbb261", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q2b", "units": [ 2 ], "type": "numerical", "sourcePages": [ 2 ], "topic": "Two-sample sigmoid forward pass and mean-MSE backprop", "id": "dl-F22-Q2b", "question": "Question (10 marks). For a three-stage sigmoid network, use samples [0.3,0.2,0.4] → 0 and [0.5,0.6,0.8] → 1. First-stage feature multipliers are [0.3,0.4,0.6]. Next receiving rows are [0.5,0.2,0.3], [0.3,0.3,0.3], and output weights [0.3,0.3]. Find both predictions, average MSE and output-weight updates after the first backward pass with η = 0.2. Assume zero biases.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 2 ], "answerRoute": "/#tee/question/tee-c15-s6", "answer_route": "/#tee/question/tee-c15-s6", "sourcePath": "sources\\S16_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Final_Exam_rTqlPdZM6a.pdf", "answerHtmlSha256": "c771af6b7fa981e873724b06211975fdef465242b1ebfa3ad783ef6ba9af427c", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q3a", "units": [ 2, 3 ], "type": "case", "sourcePages": [ 3 ], "topic": "Bias/variance, four-class activation/loss and diagnosis", "id": "dl-F22-Q3a", "question": "Question (10 marks). Explain bias, variance, underfitting and overfitting. A four-feature product-texture classifier has four grades including reject. Choose output activation/loss. Diagnose training error 15% and validation error 18% and suggest remedies.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 2, 3 ], "answerRoute": "/#tee/question/tee-c15-s7", "answer_route": "/#tee/question/tee-c15-s7", "sourcePath": "sources\\S16_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Final_Exam_rTqlPdZM6a.pdf", "answerHtmlSha256": "81b2112b84a09e47a8c3d127977165dd6f01fddc79c3b69388a51293da1503ac", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q3b", "units": [ 4 ], "type": "numerical", "sourcePages": [ 3 ], "topic": "LeNet-style parameter count", "id": "dl-F22-Q3b", "question": "Question (10 marks). Count trainable parameters of the illustrated LeNet-style digit classifier with convolution 5 × 5: 6 filters, pooling 2; convolution 5 × 5: 16 filters, pooling 2; dense 120, dense 84, softmax 10. Use the consistent 32 × 32 × 1 input that produces the drawn 28,14,10,5 spatial maps.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 4 ], "answerRoute": "/#tee/question/tee-c15-s8", "answer_route": "/#tee/question/tee-c15-s8", "sourcePath": "sources\\S16_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Final_Exam_rTqlPdZM6a.pdf", "answerHtmlSha256": "868041bbd02a5ca35cf0c37cd0429ac2b9e1d1afed6086be4c3f2507d9812b20", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q4a", "units": [ 4 ], "type": "numerical", "sourcePages": [ 3, 4 ], "topic": "Inception bottleneck computation count", "id": "dl-F22-Q4a", "question": "Question (10 marks). An Inception branch receives 28 × 28 × 192, uses 16 filters of 1 × 1 then 32 filters of 5 × 5 with spatial size preserved. Compute operations and explain 1 × 1 convolution.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 4 ], "answerRoute": "/#tee/question/tee-c15-s9", "answer_route": "/#tee/question/tee-c15-s9", "sourcePath": "sources\\S16_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Final_Exam_rTqlPdZM6a.pdf", "answerHtmlSha256": "328dbbd470c4831cf6768b9f323ac54b989e3a4956dab6e5dd7e3449f28e9fce", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q4b", "units": [ 2, 3 ], "type": "theory", "sourcePages": [ 4 ], "topic": "6-5-5-4 network; AdaGrad parameters and hyperparameters", "id": "dl-F22-Q4b", "question": "Question (10 marks). Draw a 6-input, two-hidden-layer network with 5 neurons each and 4 classes, trained by AdaGrad. Identify learned parameters and tunable hyperparameters; no input-layer weights or biases.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 2, 3 ], "answerRoute": "/#tee/question/tee-c15-s10", "answer_route": "/#tee/question/tee-c15-s10", "sourcePath": "sources\\S16_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Final_Exam_rTqlPdZM6a.pdf", "answerHtmlSha256": "4d016044aa2267b1db834cd68348590ffb3fb585ea50c0ca3771545af93dba70", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q5a", "units": [ 5 ], "type": "theory", "sourcePages": [ 4 ], "topic": "RNN model and BPTT updates", "id": "dl-F22-Q5a", "question": "Question (10 marks). Develop an RNN with hidden states and explain gradient updates using backpropagation.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 5 ], "answerRoute": "/#tee/question/tee-c15-s11", "answer_route": "/#tee/question/tee-c15-s11", "sourcePath": "sources\\S16_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Final_Exam_rTqlPdZM6a.pdf", "answerHtmlSha256": "8997569cc3bc225c8ff1f0b3899e8d2c60adc0c88e0b7eb96a5a88b6144ea3c9", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q5b", "units": [ 3 ], "type": "theory", "sourcePages": [ 4 ], "topic": "Optimization challenges and momentum", "id": "dl-F22-Q5b", "question": "Question (10 marks). Explain DL optimization challenges and the advantage of momentum over plain mini-batch gradient descent.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 3 ], "answerRoute": "/#tee/question/tee-c15-s12", "answer_route": "/#tee/question/tee-c15-s12", "sourcePath": "sources\\S16_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Final_Exam_rTqlPdZM6a.pdf", "answerHtmlSha256": "e5286b4408892cf373d2616a770904c1b647babe2584ab95ff36fa74668980f6", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q6a", "units": [ 4 ], "type": "case", "sourcePages": [ 4 ], "topic": "CNN vanishing-gradient remedies", "id": "dl-F22-Q6a", "question": "Question (10 marks). A CNN image classifier performs poorly because of vanishing gradients. Explain remedies.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 4 ], "answerRoute": "/#tee/question/tee-c15-s13", "answer_route": "/#tee/question/tee-c15-s13", "sourcePath": "sources\\S16_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Final_Exam_rTqlPdZM6a.pdf", "answerHtmlSha256": "d8a4ca7c8387b4737f919eb376a2cffd76202f371fa2255fa24bcd470f00e79f", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q6b", "units": [ 2, 3 ], "type": "case", "sourcePages": [ 4, 5 ], "topic": "Clothing classifier design and six diagnostic tasks", "id": "dl-F22-Q6b", "question": "Question (10 marks). Design a clothes classifier for 250,000 six-dimensional samples,6 classes and 2 hidden layers. Give train/validation/test split, activations/loss, actions for 15% training 20% validation error, interpret improvement to 1%/1.5%, then address 15% test error.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 2, 3 ], "answerRoute": "/#tee/question/tee-c15-s14", "answer_route": "/#tee/question/tee-c15-s14", "sourcePath": "sources\\S16_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Final_Exam_rTqlPdZM6a.pdf", "answerHtmlSha256": "0970ce0c64d7b064647980c2e024ce996a6b1e0abc6ecd011c42e8999507e333", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q7a", "units": [ 5 ], "type": "case", "sourcePages": [ 5 ], "topic": "LSTM long-sequence language model", "id": "dl-F22-Q7a", "question": "Question (10 marks). Explain with an example how LSTM handles long sequences in language modeling.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 5 ], "answerRoute": "/#tee/question/tee-c15-s15", "answer_route": "/#tee/question/tee-c15-s15", "sourcePath": "sources\\S16_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Final_Exam_rTqlPdZM6a.pdf", "answerHtmlSha256": "805324fde2708990f25f82480bf597624deef0ca78968817be2e5aaa2a9434c8", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q7b", "units": [ 2 ], "type": "theory", "sourcePages": [ 5 ], "topic": "Perceptron non-linear separability and solution", "id": "dl-F22-Q7b", "question": "Question (10 marks). Justify that single-layer perceptrons cannot classify nonlinearly separable data using an example, and suggest a solution.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 2 ], "answerRoute": "/#tee/question/tee-c15-s16", "answer_route": "/#tee/question/tee-c15-s16", "sourcePath": "sources\\S16_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Final_Exam_rTqlPdZM6a.pdf", "answerHtmlSha256": "8fbfb33a3dd1bd1045c1195221ac139506d50dc4e2ed7c697bd48626482684d9", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" } ] }, { "id": "dl-R23", "subject": "dl", "date": "2023-01-28", "type": "re-exam", "code": "R23", "sourceFile": "S15_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Re-Exam_Z8h8Mtm1AP.pdf", "sourceSha256": "6879e8b34cc84e58239206a0e49f104ba7c09c64ab215afe7393b042ce475c51", "sourcePages": 4, "visualReview": { "status": "fresh: every original page viewed 2026-10-14", "receipt": "audit/dl_nlp_all_pyq_review_20261014.md" }, "pattern": { "durationMinutes": 180, "printedMarks": 140, "attemptMarks": 100, "compulsory": [ "Q1" ], "choose": 4, "from": [ "Q2", "Q3", "Q4", "Q5", "Q6", "Q7" ] }, "questions": [ { "label": "Q1a", "units": [ 5 ], "type": "case", "sourcePages": [ 1 ], "topic": "Music-generation RNN architecture", "id": "dl-R23-Q1a", "question": "Question (4 marks). Suggest and sketch an RNN architecture to synthesize a musical tune.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 4, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 5 ], "answerRoute": "/#tee/question/tee-c16-s0", "answer_route": "/#tee/question/tee-c16-s0", "sourcePath": "sources\\S15_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Re-Exam_Z8h8Mtm1AP.pdf", "answerHtmlSha256": "64b69472b3a389a68351acf03c362480be8bd4407319772fa058a9fe8897a337", "status": "complete", "scope": "current syllabus", "retainedMarks": 4, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q1b", "units": [ 4 ], "type": "numerical", "sourcePages": [ 1 ], "topic": "Stride-2 average pooling and purpose", "id": "dl-R23-Q1b", "question": "Question (4 marks). Apply 3 × 3 average pooling with stride 2 to Matrix 1 shown below. Give the output and justify pooling. Given matrices for this question Given matrices for this question Tap the diagram to enlarge.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 4, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 4 ], "answerRoute": "/#tee/question/tee-c16-s1", "answer_route": "/#tee/question/tee-c16-s1", "sourcePath": "sources\\S15_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Re-Exam_Z8h8Mtm1AP.pdf", "answerHtmlSha256": "37ed5fe5960f8516ff35a4974cb2ba82a528c2e3cfe27e137e79409b3d9c4c73", "status": "complete", "scope": "current syllabus", "retainedMarks": 4, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q1c", "units": [ 1 ], "type": "theory", "sourcePages": [ 1 ], "topic": "DL advantages and limitations", "id": "dl-R23-Q1c", "question": "Question (4 marks). State DL advantages and limitations.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 4, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 1 ], "answerRoute": "/#tee/question/tee-c16-s2", "answer_route": "/#tee/question/tee-c16-s2", "sourcePath": "sources\\S15_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Re-Exam_Z8h8Mtm1AP.pdf", "answerHtmlSha256": "6bcfe1e1e0962baeff856d4a81343fd86a7b38da77268d38a404cccf30de4e19", "status": "complete", "scope": "current syllabus", "retainedMarks": 4, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q1d", "units": [ 3 ], "type": "theory", "sourcePages": [ 1 ], "topic": "Mini-batch size and parallelism", "id": "dl-R23-Q1d", "question": "Question (4 marks). Evaluate how mini-batch size affects optimization and parallel processing.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 4, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 3 ], "answerRoute": "/#tee/question/tee-c16-s3", "answer_route": "/#tee/question/tee-c16-s3", "sourcePath": "sources\\S15_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Re-Exam_Z8h8Mtm1AP.pdf", "answerHtmlSha256": "73c19369b245fae19c8cddfc70419605254ea4dd4d6839e4d955f151c6022e6b", "status": "complete", "scope": "current syllabus", "retainedMarks": 4, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q1e", "units": [ 2 ], "type": "numerical", "sourcePages": [ 2 ], "topic": "AND perceptron design", "id": "dl-R23-Q1e", "question": "Question (4 marks). Design a perceptron network for inputs 00 → 0,10 → 0,01 → 0,11 → 1; choose suitable weights.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 4, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 2 ], "answerRoute": "/#tee/question/tee-c16-s4", "answer_route": "/#tee/question/tee-c16-s4", "sourcePath": "sources\\S15_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Re-Exam_Z8h8Mtm1AP.pdf", "answerHtmlSha256": "44da4c40c4ed581ad0247a4c1723c5346a74baf53062412dffc43a5d7d958a9c", "status": "complete", "scope": "current syllabus", "retainedMarks": 4, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q2a", "units": [ 2, 3, 4 ], "type": "theory", "sourcePages": [ 2 ], "topic": "CNN parameters, hyperparameters and Adam", "id": "dl-R23-Q2a", "question": "Question (10 marks). A CNN with convolution, pooling, dense and output layers receives 127 × 127 images for 5 classes and trains with Adam. Identify learned parameters and tunable hyperparameters.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 2, 3, 4 ], "answerRoute": "/#tee/question/tee-c16-s5", "answer_route": "/#tee/question/tee-c16-s5", "sourcePath": "sources\\S15_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Re-Exam_Z8h8Mtm1AP.pdf", "answerHtmlSha256": "e2785bf7c34f15b2fd33b7b66e5dd120a0c563d6c6e3e3157caab99cd9d745fd", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q2b", "units": [ 2, 3 ], "type": "case", "sourcePages": [ 2 ], "topic": "ReLU, rating regression and overfitting", "id": "dl-R23-Q2b", "question": "Question (10 marks). Justify ReLU over sigmoid/tanh for deep hidden layers. For predicting brand ratings on a scale of 10, choose output activation and loss. Diagnose 2% training versus 18% validation error and suggest remedies.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 2, 3 ], "answerRoute": "/#tee/question/tee-c16-s6", "answer_route": "/#tee/question/tee-c16-s6", "sourcePath": "sources\\S15_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Re-Exam_Z8h8Mtm1AP.pdf", "answerHtmlSha256": "0101cb7c011e32a6969cdfa83b2415666c8b4f51a0fed9878cb2929862e4e8bf", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q3a", "units": [ 5 ], "type": "case", "sourcePages": [ 2 ], "topic": "RNN language model and per-token probabilities", "id": "dl-R23-Q3a", "question": "Question (10 marks). Build and train an RNN language model for “Mumbai is the financial capital of Maharashtra.” Draw the architecture and write next-token conditional probabilities at each step.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 5 ], "answerRoute": "/#tee/question/tee-c16-s7", "answer_route": "/#tee/question/tee-c16-s7", "sourcePath": "sources\\S15_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Re-Exam_Z8h8Mtm1AP.pdf", "answerHtmlSha256": "bec7c3ff6f0ecb1e17b50e31c552e9dcee798464693efbff66e5250787c62a6b", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q3b", "units": [ 2 ], "type": "numerical", "sourcePages": [ 3 ], "topic": "ReLU/sigmoid forward pass and four output-weight updates", "id": "dl-R23-Q3b", "question": "Question (10 marks). For input [0.5,0.7,0.8], first-stage multipliers [0.3,0.4,0.6], next receiving rows [0.5,0.2,0.3], [0.3,0.3,0.3], output rows [0.3,0.2], [0.2,0.3], use ReLU in the first two stages and sigmoid at outputs. Targets [0,1], zero biases, η = 0.1. Find predictions, MSE and all four hidden-to-output weight updates.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 2 ], "answerRoute": "/#tee/question/tee-c16-s8", "answer_route": "/#tee/question/tee-c16-s8", "sourcePath": "sources\\S15_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Re-Exam_Z8h8Mtm1AP.pdf", "answerHtmlSha256": "c37eeea63af29425b5bd4093f3e265443bee99909beaa03a1503b1d62a8a7cab", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q4a", "units": [ 4 ], "type": "case", "sourcePages": [ 3 ], "topic": "Small-data face-attendance transfer learning", "id": "dl-R23-Q4a", "question": "Question (10 marks). Design employee face-recognition attendance using a CNN with too few training images. Explain solutions.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 4 ], "answerRoute": "/#tee/question/tee-c16-s9", "answer_route": "/#tee/question/tee-c16-s9", "sourcePath": "sources\\S15_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Re-Exam_Z8h8Mtm1AP.pdf", "answerHtmlSha256": "01f3eded992fe4d39bc90d9b3c804e5ba161602af918fa00c9a0183df364416b", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q4b", "units": [ 4 ], "type": "numerical", "sourcePages": [ 3 ], "topic": "AlexNet parameter count", "id": "dl-R23-Q4b", "question": "Question (10 marks). Estimate parameters of the AlexNet diagram. Use Input 227 × 227 × 3. → Conv K = 11, filters = 96, stride = 4, padding = 0. → Pool K = 3, stride = 2, padding = 0. → Conv K = 5, filters = 256, stride = 1, padding = 2. → Pool K = 3, stride = 2, padding = 0. → Conv K = 3, filters = 384, stride = 1, padding = 1. → Conv K = 3, filters = 384, stride = 1, padding = 1. → Conv K = 3, filters = 256, stride = 1, padding = 1. → Pool K = 3, stride = 2, padding = 0. → Flatten. → Dense 4096 units. → Dense 4096 units. → Dense 1000 units.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 4 ], "answerRoute": "/#tee/question/tee-c16-s10", "answer_route": "/#tee/question/tee-c16-s10", "sourcePath": "sources\\S15_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Re-Exam_Z8h8Mtm1AP.pdf", "answerHtmlSha256": "3c0e841b6d5d2f77e60fc5f10e39a283730b135b41ec2546d5a3c95f1cfcd744", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q5a", "units": [ 5 ], "type": "case", "sourcePages": [ 4 ], "topic": "Vanishing-gradient RNN remedies", "id": "dl-R23-Q5a", "question": "Question (10 marks). An RNN does not improve because vanishing gradients prevent useful parameter updates. Suggest a solution and explain required architectural changes.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 5 ], "answerRoute": "/#tee/question/tee-c16-s11", "answer_route": "/#tee/question/tee-c16-s11", "sourcePath": "sources\\S15_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Re-Exam_Z8h8Mtm1AP.pdf", "answerHtmlSha256": "025573f5dc9840bc7a99d9c4b65778eed2bb13fda879a99e067a009dc2948eda", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q5b", "units": [ 2 ], "type": "mixed", "sourcePages": [ 4 ], "topic": "Fraud classifier design and parameter count", "id": "dl-R23-Q5b", "question": "Question (10 marks). Justify multiple ANN layers. Design a fraudulent/non-fraudulent transaction classifier for 10,000 samples with 4 features; choose layers/units and count parameters.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 2 ], "answerRoute": "/#tee/question/tee-c16-s12", "answer_route": "/#tee/question/tee-c16-s12", "sourcePath": "sources\\S15_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Re-Exam_Z8h8Mtm1AP.pdf", "answerHtmlSha256": "0c4c553fa367220c25b010dc565f87aef25f500dc3ee7a13c78dcaa95a06b096", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q6a", "units": [ 3 ], "type": "case", "sourcePages": [ 4 ], "topic": "Three optimization-failure scenarios", "id": "dl-R23-Q6a", "question": "Question (10 marks). Suggest and justify remedies when (1) loss does not minimize after many epochs, (2) gradients are noisy and loss oscillates, (3) loss falls then rises because updates miss a minimum.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 3 ], "answerRoute": "/#tee/question/tee-c16-s13", "answer_route": "/#tee/question/tee-c16-s13", "sourcePath": "sources\\S15_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Re-Exam_Z8h8Mtm1AP.pdf", "answerHtmlSha256": "69e59e06f073ea25d290a37563c7fecc42ebc19f66dd00b0adb3b3feba678c98", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q6b", "units": [ 4 ], "type": "numerical", "sourcePages": [ 4 ], "topic": "Convolution cost and unspecified bottleneck width", "id": "dl-R23-Q6b", "question": "Question (10 marks). A 28 × 28 × 192 input is convolved with 32 same-padded 5 × 5 filters to produce 28 × 28 × 32. Count computations. What happens if a 1 × 1 convolution precedes it? The reduction-channel count is not supplied.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 4 ], "answerRoute": "/#tee/question/tee-c16-s14", "answer_route": "/#tee/question/tee-c16-s14", "sourcePath": "sources\\S15_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Re-Exam_Z8h8Mtm1AP.pdf", "answerHtmlSha256": "6dd4bb50d7f1575146ae8d02a7d0dbf0be7e94468ce781d81965f00b1e216b2b", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "The 1x1 reduction width is not specified. Answer gives a symbolic count and an explicitly assumed numerical example." }, { "label": "Q7a", "units": [ 5 ], "type": "case", "sourcePages": [ 4 ], "topic": "LSTM sentence memory and gates", "id": "dl-R23-Q7a", "question": "Question (10 marks). Process with LSTM: “XYZ was very brave soldier. He was a part of the Kargil war and laid down his life fighting against the enemy. The president of India honored patriotic ___.” Explain the need for its gates.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 5 ], "answerRoute": "/#tee/question/tee-c16-s15", "answer_route": "/#tee/question/tee-c16-s15", "sourcePath": "sources\\S15_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Re-Exam_Z8h8Mtm1AP.pdf", "answerHtmlSha256": "2b7be2233d88a93e5e416d8c0e058d91bcbd72a6392e0e1dd0af6766ffa33c10", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" }, { "label": "Q7b", "units": [ 3 ], "type": "case", "sourcePages": [ 4 ], "topic": "Overfitting diagnosis and remedies", "id": "dl-R23-Q7b", "question": "Question (10 marks). Training performance is good but unseen-data performance degrades. Identify the issue and explain ways to improve it.", "promptStatus": "Student-facing normalized transcription, including stated assumptions; compare source for original typography.", "marks": 10, "marksStatus": "Printed whole lettered-part allocation; original marks retained even where a legacy extract is excluded.", "unit": [ 3 ], "answerRoute": "/#tee/question/tee-c16-s16", "answer_route": "/#tee/question/tee-c16-s16", "sourcePath": "sources\\S15_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Re-Exam_Z8h8Mtm1AP.pdf", "answerHtmlSha256": "55a30cf60a313870e9205088d3753c435dd13f640819d20dd2c9a01c64ea1059", "status": "complete", "scope": "current syllabus", "retainedMarks": 10, "excludedTask": null, "internalSubparts": [], "notes": "" } ] }, { "id": "dl-R25", "subject": "dl", "code": "R25", "date": "2025-02-22", "type": "re-exam", "sourceFile": "Deep Learning , Re-exam ( 2023-24, 2024-25).PDF", "sourceSha256": "5fc62eebec6981ef6b6024715dd1bd4f588b67fa8d8448a43a338a69d83c8011", "sourcePages": 6, "visualReview": { "status": "fresh: every original page visually inspected 2026-10-14", "receipt": "audit/dl_nlp_all_pyq_review_20261014.md" }, "pattern": { "durationMinutes": 180, "printedMarks": 140, "attemptMarks": 100, "compulsory": [ "Q1" ], "choose": 4, "from": [ "Q2", "Q3", "Q4", "Q5", "Q6", "Q7" ], "note": "Combined-year archive filename contains one six-page exam, not two independent papers." }, "questions": [ { "id": "dl-R25-Q1a", "label": "Q1a", "question": "Compare ML and DL. Choose architectures for five-class diabetic-retinopathy retinal images, alphabet-image recognition, and weather prediction from N days of history.", "marks": 5, "marksStatus": "Printed allocation", "unit": [ 1, 4, 5 ], "units": [ 1, 4, 5 ], "topic": "ML versus DL and architecture choice", "type": "case", "sourcePages": [ 1 ], "answerRoute": "/#tee/question/tee-c21-s0", "status": "complete", "scope": "current syllabus", "notes": "", "internalSubparts": [], "answerHtmlSha256": "39bd0906c21cbd57978a359bcb6d818e7bcf4fc72a1f1b4023d4a492cf230b9a", "answerTaskChecklist": [ { "task": "Compare ML and DL", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "Retinopathy architecture", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "Alphabet-image architecture", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "Weather-sequence architecture", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" } ] }, { "id": "dl-R25-Q1b", "label": "Q1b", "question": "Choose best and worst models from training/validation errors: M1=0.9/1.2; M2=0.3/0.6; M3=1.5/0.5; M4=1.2/1.1. Explain the reasoning and remedies.", "marks": 5, "marksStatus": "Printed allocation", "unit": [ 3 ], "units": [ 3 ], "topic": "Training and validation errors", "type": "case", "sourcePages": [ 1 ], "answerRoute": "/#tee/question/tee-c21-s1", "status": "complete", "scope": "current syllabus", "notes": "", "internalSubparts": [], "answerHtmlSha256": "9a960afa29c7cd1bcd1b19e770dc19b1a761157af8b4d1c9dc86389336e31bd7", "answerTaskChecklist": [ { "task": "Best and worst from validation errors", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "Interpret training versus validation", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "Explain appropriate remedies and limits", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" } ] }, { "id": "dl-R25-Q1c", "label": "Q1c", "question": "Explain why CNNs use a hierarchy of convolutional layers. Apply the printed3×3kernel to the5×5input with stride2.", "marks": 5, "marksStatus": "Printed allocation", "unit": [ 4 ], "units": [ 4 ], "topic": "CNN hierarchy and stride-two convolution", "type": "mixed", "sourcePages": [ 1 ], "answerRoute": "/#tee/question/tee-c21-s2", "status": "complete with explicit source ambiguity", "scope": "current syllabus", "notes": "Padding is not specified; valid/no-padding is declared. CNN cross-correlation convention is used without kernel reversal.", "internalSubparts": [], "answerHtmlSha256": "9a21c3cc32f7f1a122f796326fd8f66dae433cba341ed086e667d072192a9c68", "answerTaskChecklist": [ { "task": "CNN feature hierarchy", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "Four stride-two convolution cells", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" } ] }, { "id": "dl-R25-Q1d", "label": "Q1d", "question": "Compare BPTT with feedforward backpropagation and state the strengths of RNNs.", "marks": 5, "marksStatus": "Printed allocation", "unit": [ 5 ], "units": [ 5 ], "topic": "BPTT and recurrent strengths", "type": "theory", "sourcePages": [ 1, 2 ], "answerRoute": "/#tee/question/tee-c21-s3", "status": "complete", "scope": "current syllabus", "notes": "", "internalSubparts": [], "answerHtmlSha256": "99a6ee592e1a85fe5685e25dbae08c21c9c17ee5c9d214ab0897bbf84a4dbf49", "answerTaskChecklist": [ { "task": "BPTT versus feedforward backprop", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "RNN strengths", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" } ] }, { "id": "dl-R25-Q2a", "label": "Q2a", "question": "Explain why activation functions are needed and sketch ReLU. Is sigmoid appropriate for an unscaled house-price output? Compute softmax([1.2,3.4,0.9]) and sigmoid(2.5).", "marks": 10, "marksStatus": "Printed allocation", "unit": [ 2 ], "units": [ 2 ], "topic": "Activation, ReLU and numeric outputs", "type": "mixed", "sourcePages": [ 2 ], "answerRoute": "/#tee/question/tee-c21-s4", "status": "complete", "scope": "current syllabus", "notes": "", "internalSubparts": [], "answerHtmlSha256": "7335805fef273aa8076b0d1d1003d84e5ad7da2ed86879215034f1de12da1e81", "answerTaskChecklist": [ { "task": "Need for activation", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "ReLU sketch", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "House-price output suitability", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "Three-class softmax", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "Sigmoid2.5", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" } ] }, { "id": "dl-R25-Q2b", "label": "Q2b", "question": "Explain vanishing/exploding gradients in RNNs and the gates of a GRU.", "marks": 10, "marksStatus": "Printed allocation", "unit": [ 5 ], "units": [ 5 ], "topic": "Vanishing/exploding gradients and GRU", "type": "theory", "sourcePages": [ 2 ], "answerRoute": "/#tee/question/tee-c21-s5", "status": "complete", "scope": "current syllabus", "notes": "", "internalSubparts": [], "answerHtmlSha256": "966df376b39457a7f00b6651018198b12231c6a70e31b410610df7b6762c2ef6", "answerTaskChecklist": [ { "task": "Vanishing and exploding gradients", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "GRU gates and recurrence", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" } ] }, { "id": "dl-R25-Q3a", "label": "Q3a", "question": "For the printed network, inputs=[0.1,0.2,0.3,0.4], targets=[0.01,0.99], w1…w18=[0.10,0.15,…,0.95] and both layer biases=0.60. Find total forward error, then drop hidden neuron H12 and compare errors.", "marks": 10, "marksStatus": "Printed allocation", "unit": [ 2, 3 ], "units": [ 2, 3 ], "topic": "4–3–2 forward pass and dropout", "type": "numerical", "sourcePages": [ 3 ], "answerRoute": "/#tee/question/tee-c21-s6", "status": "complete with explicit source ambiguity", "scope": "current syllabus", "notes": "Activation/loss/dropout scaling missing from original; result is conditional on explicit standard classroom assumptions.", "internalSubparts": [], "answerHtmlSha256": "7a82bbfd26a71ef0e85f0a57befee3f263e5f13d5a6691db8f334f3074696c3c", "answerTaskChecklist": [ { "task": "Original forward outputs and total loss", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "H12 dropout forward outputs and loss", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "Compare losses", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" } ] }, { "id": "dl-R25-Q3b", "label": "Q3b", "question": "Explain(i)parameter sharing and sparse connections,(ii)5×5versus3×3kernel trade-offs,(iii)ReLU’s effect on learning,(iv)whether CNNs can work without pooling.", "marks": 10, "marksStatus": "Printed allocation", "unit": [ 4 ], "units": [ 4 ], "topic": "Four CNN design explanations", "type": "theory", "sourcePages": [ 3 ], "answerRoute": "/#tee/question/tee-c21-s7", "status": "complete", "scope": "current syllabus", "notes": "", "internalSubparts": [], "answerHtmlSha256": "b8c573c124ec9490edb233ebf1c7f801343de064fc55450b6f07d8b2b4e77913", "answerTaskChecklist": [ { "task": "Parameter sharing and sparse connectivity", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "5x5 versus3x3 tradeoff", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "ReLU learning effect", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "CNN without pooling", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" } ] }, { "id": "dl-R25-Q4a", "label": "Q4a", "question": "For input64×64×3: conv3×3,32filters,P1,S1; maxpool2×2,S2; conv3×3,64,P1,S1; pool2,S2; conv3×3,128,P1,S1; pool2,S2; FC256; softmax43. Find each shape and all weights/biases.", "marks": 10, "marksStatus": "Printed allocation", "unit": [ 4 ], "units": [ 4 ], "topic": "64×64 CNN shape and parameter table", "type": "numerical", "sourcePages": [ 4 ], "answerRoute": "/#tee/question/tee-c21-s8", "status": "complete", "scope": "current syllabus", "notes": "", "internalSubparts": [], "answerHtmlSha256": "07011da7583ebf5064f189f3d4e1c9868b87d433d048d54aa502dbd45a0a53cc", "answerTaskChecklist": [ { "task": "All layer shapes", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "All weights and biases", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "Total parameters", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" } ] }, { "id": "dl-R25-Q4b", "label": "Q4b", "question": "Implement AND with bipolar inputs/targets and learning rate1.", "marks": 10, "marksStatus": "Printed allocation", "unit": [ 2 ], "units": [ 2 ], "topic": "Bipolar AND perceptron", "type": "numerical", "sourcePages": [ 4 ], "answerRoute": "/#tee/question/tee-c21-s9", "status": "complete with explicit source ambiguity", "scope": "current syllabus", "notes": "Initialization, order and threshold convention explicitly supplied as assumptions; multiple equivalent separating weights exist.", "internalSubparts": [], "answerHtmlSha256": "49dface5cc1a4e59a42cafeac1735326174de933a18f3e2724b5c4f9b1031e7c", "answerTaskChecklist": [ { "task": "Bipolar truth table", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "Declared initialization and tie rule", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "Per-example updates", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "Final gate verification", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" } ] }, { "id": "dl-R25-Q5a", "label": "Q5a", "question": "Describe skip connections and their benefit in ResNet. Explain a1×1convolution’s role in Inception with an example.", "marks": 10, "marksStatus": "Printed allocation", "unit": [ 4 ], "units": [ 4 ], "topic": "ResNet shortcut and Inception1×1", "type": "theory", "sourcePages": [ 4 ], "answerRoute": "/#tee/question/tee-c21-s10", "status": "complete", "scope": "current syllabus", "notes": "", "internalSubparts": [], "answerHtmlSha256": "155375f24fa077ab431a9a6c23b113eae47435a70487be5d42e7272bd7e5ebf0", "answerTaskChecklist": [ { "task": "ResNet skip connection and benefit", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "Inception1x1 role and example", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" } ] }, { "id": "dl-R25-Q5b", "label": "Q5b", "question": "Sketch a one-hidden-layer RNN and state its hidden-state formula. Compute hₜ for Wxh=[[0.2,0.4],[0.1,0.3]], Whh=[[0.5,0.6],[0.3,0.4]], b=[0.1,0.2], x=[1,0], previoush=[0,1], ReLU.", "marks": 10, "marksStatus": "Printed allocation", "unit": [ 5 ], "units": [ 5 ], "topic": "ReLU RNN state with declared orientation", "type": "numerical", "sourcePages": [ 4 ], "answerRoute": "/#tee/question/tee-c21-s11", "status": "complete with explicit source ambiguity", "scope": "current syllabus", "notes": "Square matrices and printed horizontal vectors leave orientation ambiguous; both possible readings are explained, main answer follows row-vector form.", "internalSubparts": [], "answerHtmlSha256": "82ba8bde585c5a559639b44011f13e8f9d7999aad8aa073e15f34c2500cc9770", "answerTaskChecklist": [ { "task": "RNN diagram and hidden-state formula", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "Exact input and recurrent products", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "ReLU", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "Orientation qualification", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" } ] }, { "id": "dl-R25-Q6a", "label": "Q6a", "question": "For inputs[1,0,1], weights w1…w8=[0.2,−0.3,0.4,0.1,−0.5,0.2,−0.3,−0.2], biases[−0.4,0.2,0.1], target1 and sigmoid at all neurons: find prediction,error,and update w7,w8 with η=0.9.", "marks": 10, "marksStatus": "Printed allocation", "unit": [ 2 ], "units": [ 2 ], "topic": "3–2–1 forward pass and update w7,w8", "type": "numerical", "sourcePages": [ 5 ], "answerRoute": "/#tee/question/tee-c21-s12", "status": "complete with explicit source ambiguity", "scope": "current syllabus", "notes": "Original does not specify loss scaling; half-SSE is declared and alternatives distinguished.", "internalSubparts": [], "answerHtmlSha256": "e4f28b9c0c78cf9c69a269d9f2b5a84fd1c08f6feab6381976a288c44a62e963", "answerTaskChecklist": [ { "task": "Forward prediction", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "Residual and error scaling", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "Update w7", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "Update w8", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" } ] }, { "id": "dl-R25-Q6b", "label": "Q6b", "question": "A healthcare model trained with AdaGrad becomes stagnant after many epochs. Discuss AdaGrad and RMSProp in detail and their fundamental differences.", "marks": 10, "marksStatus": "Printed allocation", "unit": [ 3 ], "units": [ 3 ], "topic": "AdaGrad stagnation versus RMSProp", "type": "case", "sourcePages": [ 5 ], "answerRoute": "/#tee/question/tee-c21-s13", "status": "complete", "scope": "current syllabus", "notes": "", "internalSubparts": [], "answerHtmlSha256": "0ddcb434b1bce675299a75d83854c4a9d7cab3b0a1050c0f3666cd178ce5aa13", "answerTaskChecklist": [ { "task": "AdaGrad mechanism and formula", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "RMSProp mechanism and formula", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "Accumulation versus decay", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "Stagnation interpretation", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" } ] }, { "id": "dl-R25-Q7a", "label": "Q7a", "question": "Choose a deep architecture for long social-media/customer-review sequences classified positive,negative or neutral. Explain its operation with equations.", "marks": 10, "marksStatus": "Printed allocation", "unit": [ 5 ], "units": [ 5 ], "topic": "Long-text sentiment architecture", "type": "case", "sourcePages": [ 5, 6 ], "answerRoute": "/#tee/question/tee-c21-s14", "status": "complete", "scope": "current syllabus", "notes": "", "internalSubparts": [], "answerHtmlSha256": "23fb755b67cdbec75d120740c1204e6fe2384d3be367e044f64a854408361667", "answerTaskChecklist": [ { "task": "Choose long-sequence sentiment architecture", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "Architecture operation and gate equations", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "Three-class head", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" } ] }, { "id": "dl-R25-Q7b", "label": "Q7b", "question": "Explain transfer learning and its benefits with suitable examples.", "marks": 5, "marksStatus": "Printed allocation", "unit": [ 4 ], "units": [ 4 ], "topic": "Transfer learning", "type": "theory", "sourcePages": [ 6 ], "answerRoute": "/#tee/question/tee-c21-s15", "status": "complete", "scope": "current syllabus", "notes": "", "internalSubparts": [], "answerHtmlSha256": "1606f165aca351e35f588cb501f319fe3635b68831d562e51352605770c982b8", "answerTaskChecklist": [ { "task": "Transfer learning definition", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "Benefits", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "Examples", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" } ] }, { "id": "dl-R25-Q7c", "label": "Q7c", "question": "Explain the importance of learning rate. Give and explain one classification loss and one regression loss.", "marks": 5, "marksStatus": "Printed allocation", "unit": [ 2, 3 ], "units": [ 2, 3 ], "topic": "Learning rate and two losses", "type": "theory", "sourcePages": [ 6 ], "answerRoute": "/#tee/question/tee-c21-s16", "status": "complete", "scope": "current syllabus", "notes": "", "internalSubparts": [], "answerHtmlSha256": "8d88add1a8b2ada43a096a73eb66310be3015d10095f20f6b9555445f52ade9a", "answerTaskChecklist": [ { "task": "Learning-rate importance", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "Classification loss", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" }, { "task": "Regression loss", "status": "addressed; source ambiguity stated where applicable", "marks": null, "marksRule": "shares parent allocation; do not sum separately" } ] } ] } ], "catalogueCandidates": [ { "id": "dl-01", "canonicalPaper": "dl-F25", "sourceSha256": "ba5e45671bd189890d9d8bd2d2ee6a3c0f8e11f49e69ecb84a9e2dd0b5fab2fc", "sourceFile": "S18_B_Tech__Deep_Learning__2025-26_Final_Exam_wH9YsCG0PI.pdf", "aliases": [ "02_Question_Papers/Legacy/B TECH/3RD YEAR/ARTIFICAL INTELLIGENCE/SEM-V/DEEP LEARNING/B Tech, Deep Learning, 2025-26 Final Exam.pdf", "02_Question_Papers/Legacy/MBA TECH/3RD YEAR/ARTIFICIAL INTELLIGENCE/Deep Learning/Deep Learning, 2025-26 Final Exam.pdf", "02_Question_Papers/Legacy/MBA TECH/3RD YEAR/ARTIFICIAL INTELLIGENCE/Deep Learning/Deep Learning.pdf" ], "status": "restored and all pages visually checked; duplicate scan" }, { "id": "dl-02", "canonicalPaper": "dl-R24", "sourceSha256": "3df2c1997c7790a13ab181eae53997097bc628f2c0949eaf6810f8324fa07f23", "sourceFile": "S14_Deep_Learning___Re-exam_2023-24_oxnB58Pwaa.pdf", "aliases": [ "02_Question_Papers/Legacy/B TECH/3RD YEAR/ARTIFICAL INTELLIGENCE/SEM-V/DEEP LEARNING/Deep Learning , Re-exam 2023-24.pdf", "02_Question_Papers/Legacy/MBA TECH/3RD YEAR/ARTIFICIAL INTELLIGENCE/Deep Learning/Deep Learning , Re-exam 2023-24.pdf" ], "status": "original pages visually checked; canonical examination counted once" }, { "id": "dl-03", "canonicalPaper": "dl-F22", "sourceSha256": "39d3f520183f42336d1a7671aa946f3a6555484b1c5f6d7b8cdcb513efbda374", "sourceFile": "S16_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Final_Exam_rTqlPdZM6a.pdf", "aliases": [ "02_Question_Papers/Legacy/B TECH/3RD YEAR/ARTIFICAL INTELLIGENCE/SEM-V/DEEP LEARNING/Deep Learning, 3rd Year, Sem-V, Year 2022-23, Final Exam.pdf" ], "status": "original pages visually checked; canonical examination counted once" }, { "id": "dl-04", "canonicalPaper": "dl-R23", "sourceSha256": "6879e8b34cc84e58239206a0e49f104ba7c09c64ab215afe7393b042ce475c51", "sourceFile": "S15_Deep_Learning__3rd_Year__Sem-V__Year_2022-23__Re-Exam_Z8h8Mtm1AP.pdf", "aliases": [ "02_Question_Papers/Legacy/B TECH/3RD YEAR/ARTIFICAL INTELLIGENCE/SEM-V/DEEP LEARNING/Deep Learning, 3rd Year, Sem-V, Year 2022-23, Re-Exam.pdf" ], "status": "original pages visually checked; canonical examination counted once" }, { "id": "dl-05", "canonicalPaper": "dl-F23", "sourceSha256": "017671e0097f1bfe6baea738b916dc9d38d3a78006c85831654fd4d4e7109e41", "sourceFile": "S17_Deep_Learning__Final_exam_2023-24_EyPdFZ3lIK.pdf", "aliases": [ "02_Question_Papers/Legacy/B TECH/3RD YEAR/ARTIFICAL INTELLIGENCE/SEM-V/DEEP LEARNING/Deep Learning, Final exam,2023-24.pdf", "02_Question_Papers/Legacy/MBA TECH/3RD YEAR/ARTIFICIAL INTELLIGENCE/Deep Learning/Deep Learning, Final exam,2023-24.pdf" ], "status": "original pages visually checked; canonical examination counted once" }, { "id": "dl-12", "canonicalPaper": "dl-R25", "sourceSha256": "5fc62eebec6981ef6b6024715dd1bd4f588b67fa8d8448a43a338a69d83c8011", "sourceFile": "Deep Learning , Re-exam ( 2023-24, 2024-25).PDF", "aliases": [ "02_Question_Papers/Legacy/MBA TECH/3RD YEAR/ARTIFICIAL INTELLIGENCE/Deep Learning/Deep Learning , Re-exam ( 2023-24, 2024-25).PDF" ], "status": "original pages visually checked; canonical examination counted once" } ], "inventoryScope": "All AI third-year same-course PDF entries in the supplied complete library manifest plus the original local final/re-exam set; programme aliases and duplicate scans deduplicated. This is not a claim that no other unprovided paper exists." }