What a neural network actually computes, why the nonlinearity is the whole trick, and how the weights learn.
Python
PyTorch
NumPyThis is the real simulation from the module. No account needed.
A suggestion, never a lock — you can start here.
Every lesson's title and summary is public. The lessons themselves open once you own the track. The lesson marked free can be read in full right now, no account needed.
Every lesson in Introduction to Deep Learning & Neural NetworksA neuron is a weighted sum plus one bending function. Stack a few and a straight line becomes whatever boundary you need.
Why the sigmoid has all but vanished from hidden layers, and what replaced it.
The chain rule, computation graphs, and why one backward pass is enough for millions of weights.
From tensors to a trained model: forward, loss, backward, step — and how to read the curves.
Optimizers, learning-rate schedules, initialisation, normalisation and regularisation — what makes training actually converge.
Tensors, autograd, nn.Module, DataLoader and GPU training — enough PyTorch to work on your own.
Convolution, pooling and CNN architectures — why weight sharing makes images learnable.
From RNNs and LSTMs to self-attention — the architecture behind every modern LLM.
