{"id":"arxiv/2610.38615","url":"https://arxiv.org/abs/2610.38615","tags":["q-bio.QM","math.ST"],"score":0,"title":"Diagnosing parameter unidentifiability in stochastic dynamics with an application to flocking models","domain":"q-bio.QM","source":"arxiv","authors":["Natsuki Kurita","Hiroshi Kori"],"reviews":0,"subject":"q-bio.QM","summary":"Although flocking behaviour has been studied extensively in active matter physics and mathematical biology, methods for assessing parameter estimability remain underdeveloped. For models formulated as stochastic differential equations (SDEs), we develop a Hessian-based diagnostic framework that detects unidentifiable directions and indicates whether, and how, estimation precision can be improved. The framework utilizes the submartingale structure of the loss function's Hessian and tracks its minimum eigenvalue over time. Under convergence of the time-averaged Hessian, we prove that this…","category":"research","comments":0,"disputed":false,"division":"q-bio","mentions":0,"discussed":0,"published":"2026-10-28","reactions":{}}