Trine recovers state-dependent noise in dynamical systems by learning sign before magnitude
bravo_abad · x · 2026-09-24
In many scientific systems, noise strength varies with the system's state, and that variation carries useful physics or biology. The Trine method avoids estimating hidden noise all at once: it first learns the smooth deterministic dynamics with a Gaussian process, then estimates only the sign of stochastic fluctuations from the residuals, uses a structured kernel to reconstruct noise realizations from those signs, and finally regresses how noise amplitude changes across state space. Learning direction before magnitude is statistically easier than recovering full amplitude, and the authors show this intermediate step pays off.
More from Research
- Claude Opus 5.5 and GPT-6 Sol land on Biomni Lab research platform — KexinHuang5 · 2026-09-24
- Arc Institute unveils Minerva, a genome language model for predicting RNA interactions — BrianHie · 2026-09-24
- JevK5 open decision model ranks #2 of 76 on JevBench, outputs calibrated probabilities in ~13ms — airesearch12 · 2026-09-24
- Schmidt Sciences commits up to $20M to eight 2026 Polymath researchers — arjunrajlab · 2026-09-24
- World Labs releases Atlas, an omni world model native to text, images, video, and 3D — theworldlabs · 2026-09-24
- zeta(5) Likely Proven Irrational, Author Says Astra Verified the Argument — aran_nayebi · 2026-09-24