01 — MLP and Representation Learning

Fully connected feedforward networks — the simplest deep learning building block and the foundation for understanding depth, activation functions, and backpropagation.

Notes

• Multi-Layer Perceptron (MLP) — architecture, activation functions, forward and backward pass, depth vs width trade-offs
• Deep Learning Training Patterns (PyTorch) — MLP, CNN, RNN training loops with PyTorch, early stopping, validation loop, model serialisation (in 08_implementations)

Links

← 04 Deep Learning → 02 — Convolutional Networks

• Deep Learning Theory — backpropagation, activations
• Training and Regularization