From the mathematics of tensors to activation functions and backpropagation — the foundations of neural networks, taught by example.
Lecture notes by Dr. Abdulkarim Albanna. References: Goodfellow, Bengio & Courville; Trask; Chollet; Nielsen.
One concept per module, building from the math a network runs on up to the algorithm that trains it. Each module is a written tutorial with worked examples, figures, and exercises with full solutions.
Every concept in these notes can be implemented and tested in PyTorch on the companion site — “like LeetCode, but for tensors”: instant feedback, reference solutions, no GPU needed.
Open TorchCodeDeep learning goes well beyond a single neuron. These modules are on the way — same format, same notes.
This is the foundations unit of a Deep Learning course, built directly from Dr. Abdulkarim Albanna's lecture notes and the standard references (Goodfellow, Bengio & Courville — Deep Learning; Trask — Grokking Deep Learning; Chollet — Deep Learning with Python; Nielsen). Every module follows the same rhythm: learn the concept in plain English, walk a worked example with real numbers, then test yourself with an exercise and its solution. We start from the linear algebra and numerical computation a network runs on, move through activation functions and a hand-built XOR network, and finish with a complete forward and backward pass of backpropagation.