Petar Veličković on Categorical Deep Learning: An Algebraic Theory of Architectures
burny_tech · x · 2026-09-03
Petar Veličković (DeepMind, author of GAT) presented "Categorical Deep Learning: An Algebraic Theory of Architectures", which builds a unified category-theoretic framework to formally characterize and guide neural architecture design.
Related event: DeepMind's Petar Veličković Lectures on Categorical Deep Learning(2 posts)→
More from Research
- NeurIPS 2026 medical agents workshop recruits volunteer reviewers — mdredze · 2026-09-03
- Pyr Opens Mouse Hippocampus CA3 Connectome Dataset with 2,000 Neurons, 36M Synapses — burny_tech · 2026-09-03
- Open-Source Demo: AI Agent Reverse-Engineers Black-Box Golf Physics in Closed Loop — burny_tech · 2026-09-03
- Deep Sets Revisited: The Theorem That Let Neural Nets Ignore Input Order — burny_tech · 2026-09-03
- DeepONet revisited: single hidden layer nets can approximate any nonlinear operator — burny_tech · 2026-09-03
- ExecRetrieval benchmark shows code embedders rank buggy near-clones above correct code — kennesaw · 2026-09-03