Neural Networks
Build intuition for how neural networks learn — from a single neuron all the way to backpropagation.
7 lessons
01What is a Neuron?
The building block of every neural network — what it computes, why it needs an activation, and what one neuron can (and can't) learn.
Gradient Descent
How a network learns — following the slope of the loss surface downhill, and the optimizers that speed it up.
Layers & the Forward Pass
Stacking neurons into layers turns dot products into matrix multiplications — and stacking layers lets simple features compose into complex ones.
XOR, MLPs & Softmax
Why a single perceptron can't solve XOR, how a hidden layer fixes it, and how softmax turns raw scores into probabilities.
BatchNorm & Dropout
The two training stabilizers every deep network uses — why internal covariate shift kills training, and how dropout acts as an ensemble of networks.
Weight Initialization & Training Dynamics
Getting training to start at all — why initialization scale decides whether signals vanish or explode, Xavier and He initialization, and the residual connections, normalization, and warmup that keep deep nets stable.
Neural Networks Quiz
Test your understanding of neurons, activations, forward passes, and backpropagation.