NEURAL NETWORKS FROM SCRATCH

LEARNING OUTCOME

Building a deep neural network with bare Python and NumPy.

Content

Hands-on workshop on: supervised learning, binary classification models, fully connected feedforward neural networks, gradient descent, backpropagation, the training function, bias / variance tradeoff, regularization, optimization algorithms.

This idea has been observed in the wild many times already: Andrej Karpathy once said that one of the best way to learn deep learning is to build your own library. Marco will provide his own walkthrough: fast-paced but easy to follow, dense but hopefully fun.