'The Principles of Deep Learning Theory' by Roberts & Yaida published by Cambridge
FrnkNlsn · x · 2026-10-29
"The Principles of Deep Learning Theory" by Daniel A. Roberts (MIT) and Sho Yaida (Meta AI), with contributions from Boris Hanin, is out from Cambridge University Press (May 2022, 472 pages, $89 digital). Borrowing effective-theory methods from theoretical physics, the textbook builds a rigorous yet pedagogical framework for how realistic deep neural networks actually work, spanning physics, CS, and statistical physics. A free arXiv preprint is available.
More from Research
- Michael Hahn at BlackboxNLP: Turning Transformer Computations into Programs and Automata — hanjie_chen · 2026-10-29
- BlackboxNLP Orals: Emergent Misalignment Data Attribution, Phonotactics, In-Context Circuits — hanjie_chen · 2026-10-29
- New paper: agentic economies for scientific discovery—allocating scarce experiment capacity — sebkrier · 2026-10-29
- AI audit of 138 economics papers finds 2,035 errors; 97% of directional errors inflate claims — eigenrobot · 2026-10-29
- Researchers steer four neurons in Qwen3.5-4B to turn it into a catgirl named Meecot — burny_tech · 2026-10-29
- Is RAG dead? STAIR's structure-aware retrieval sparks debate on post-chunking architectures — Money-Designer-9724 · 2026-10-29