A Simple ODE Predicts MLP Training Dynamics, New Work Revives ML Basic Science
joshalbrecht · x · 2026-11-17
A new research share from learningmech shows that a simple ordinary differential equation can predict the training dynamics of MLPs, aimed at people working on the fundamental science of ML training. Researcher joshalbrecht amplified it, musing that it's nice to see people still doing ML as science.
More from Research
- Latent feedback cuts Transformer training compute by up to 46% with 4 lines of code — IFM · 2026-11-17
- PaperCritic: RL-trained AI reviewers beat frontier models at judging paper soundness — CarnegieMellonU · 2026-11-17
- LoopFill: training-free speculative decoding speeds up looped Transformers — UIUC-CS · 2026-11-17
- SPILLWAY: deterministic defense stops all templated prompt injections on tool-using agents — saintmachines · 2026-11-17
- SJTU's Calpha folds proteins from a general-purpose LLM's internal residue states — SJTU · 2026-11-17
- AllocFlow: end-to-end flow matching turns raw market data into portfolio allocations — Annan Wang · 2026-11-17