Deep Learning

'Deep Learning' by Goodfellow, Bengio, and Courville is the canonical reference work in deep learning. It covers everything from basic neural networks to advanced architectures, and most importantly: it explains not just what works but why.

For those who have read the attention-and-hallucinations article, this work is valuable for understanding the broader context of transformers, backpropagation, and the mathematical principles behind them. Yoshua Bengio is one of three authors behind the attention breakthrough, and his perspective runs through the entire book.