Neural net distilled into 26-node equation predicts planetary instability, beating analytic formulas
MilesCranmer · x · 2026-11-16
Researchers from Cornell, MIT and Cambridge distilled a neural network into closed-form algebraic equations that predict when multi-planet systems go dynamically unstable. Training on N-body simulations then applying symbolic regression yields a 26-node equation with 1.34–1.52 orders of magnitude accuracy on 3-planet systems, beating the Petit+2020 analytic formula (2.43–3.13 dex), and it zero-shot extrapolates to 5-planet systems. The two-stage pipeline shows symbolic learning can extract scientific insight from black-box networks while evaluating faster than both net inference and direct simulation.
More from Research
- Chollet: reasoning chain generation is search, not interpolation — teortaxesTex · 2026-11-16
- Factorised output heads let tiny LMs use a 131k vocab at equal training FLOPs — BLCNYY · 2026-11-16
- Terence Tao: math has moved from proof scarcity to proof abundance — haider1 · 2026-11-16
- Wrong AI advice hurts doctors' judgment more than right advice helps: study of 16,641 decisions — ValerioCapraro · 2026-11-16
- Free online book 'The Science of Computing' links PDEs, numerics, and cache performance — burkov · 2026-11-16
- Controlled speech experiments: joint training from scratch always beats separate pretraining — giffmana · 2026-11-16