Deep Learning · MT2 · chapter

Neural-network foundations

Technical definitions, neuron models, activations, losses, forward propagation and backpropagation. These foundations support the MT2 calculations.

1. What does learning from data mean
2. Classification regression and learning types
3. Training validation and test data
4. Parameters and hyperparameters
5. What one artificial neuron does
6. Activation functions and how to calculate them
7. Drill 8 Activation contrast
8. Choosing a loss
9. Drill 2 Half sum versus mean
10. Drill 7 Sigmoid with binary cross entropy
11. Drill 9 Softmax and true class loss
12. Cross-entropy for probability distributions
13. Forward propagation through multiple neurons
14. Your first complete training step
15. Complete sigmoid training example
16. Drill 10 Sigmoid neuron gradient
17. Backpropagation with one hidden neuron
18. More outputs and vectorized backpropagation
19. Drill 11 Full ReLU backpropagation
20. Drill 12 Two output sigmoid backpropagation
21. Drill 13 A negative ReLU blocks this path
22. Counting parameters in a dense network
23. Teacher vectorized ReLU example
24. Teams photos full sigmoid example
25. MT1A Q2
26. MT1A Q3
27. MT1A Q4
28. MT1B Q3
29. MT1B Q4