Neurons, layers, forward propagation and a training loop, before the deep learning track takes it apart.
7h · 9 lessons · 1 challenge
Before this: Overfitting, Regularization & Model Selection
A neural network is a stack of the logistic units you already built. This module assembles them: neurons and layers, forward propagation by hand and vectorized, then the training loop that fits them.
It is deliberately the first pass. The Deep Learning track rebuilds all of this from scratch and takes the backward pass apart properly; this is the on-ramp that makes that possible.
A target function and a budget of neurons. Assemble the smallest network that fits it.
Worth 350 XP.
