Neural Networks and Deep Learning

Michael Nielsen · 2015 · Beginner

Nielsen builds a working digit classifier from nothing, then spends six chapters explaining why every piece of it works. It is the rare book that derives backpropagation slowly enough to follow and honestly enough to trust — including a whole chapter on why deep networks are hard to train at all.

Start reading6 chapters · 83,965 words · 6h 22m

Chapters

Using neural nets to recognize handwritten digits1h 12m How the backpropagation algorithm works38m Improving the way neural networks learn2h 12m A visual proof that neural nets can compute any function31m Why are deep neural networks hard to train?26m Deep learning1h 23m

Pairs with these lessons

Neural Networks Backpropagation Normalization & Initialization Regularization