Current browse context:
stat.ML
Change to browse by:
References & Citations
Statistics > Machine Learning
Title: Sparsistent Model Discovery
(Submitted on 22 Jun 2021 (v1), last revised 5 Oct 2021 (this version, v2))
Abstract: Discovering the partial differential equations underlying spatio-temporal datasets from very limited and highly noisy observations is of paramount interest in many scientific fields. However, it remains an open question to know when model discovery algorithms based on sparse regression can actually recover the underlying physical processes. In this work, we show the design matrices used to infer the equations by sparse regression can violate the irrepresentability condition (IRC) of the Lasso, even when derived from analytical PDE solutions (i.e. without additional noise). Sparse regression techniques which can recover the true underlying model under violated IRC conditions are therefore required, leading to the introduction of the randomised adaptive Lasso. We show once the latter is integrated within the deep learning model discovery framework DeepMod, a wide variety of nonlinear and chaotic canonical PDEs can be recovered: (1) up to $\mathcal{O}(2)$ higher noise-to-sample ratios than state-of-the-art algorithms, (2) with a single set of hyperparameters, which paves the road towards truly automated model discovery.
Submission history
From: Georges Tod [view email][v1] Tue, 22 Jun 2021 17:19:36 GMT (3024kb,D)
[v2] Tue, 5 Oct 2021 15:40:28 GMT (2228kb,D)
Link back to: arXiv, form interface, contact.