The Little Book of Deep Learning

Livro digital

Título:
The Little Book of Deep Learning

Autor:
François Fleuret

Categoria:
Tecnologia > IA

Doador:
Raffaello D. N.

Sinopse:
If you want a compact path through deep learning without losing the logic of the field, this book starts exactly where the subject becomes real: Foundations, then Deep Models, then Applications. François Fleuret uses that structure to move from learning from data, underfitting and overfitting, and gradient descent to the machinery that makes modern models work. The table of contents shows a careful progression: efficient computation with GPUs and tensors, training protocols and backpropagation, then model components such as activation functions, dropout, normalization, skip connections, attention layers, token embedding, and positional encoding. It then connects those parts to architectures like MLPs, convolutional networks, and attention models, before turning to prediction tasks and synthesis. The result is a concise but serious guide that feels both historical and current, since it reaches from the Neocognitron to prompt engineering, quantization, adapters, and model merging. Readers get a clear conceptual map of deep learning, plus enough practical detail to understand how core architectures and modern workflows fit together.