Sparse Learning of Dynamical Systems in RKHS: An Operator-Theoretic Approach

Authors: Boya Hou, Sina Sanjari, Nathan Dahlin, Subhonmesh Bose, Umesh Vaidya

ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this work, we present a method for sparse learning of transfer operators from βmixing stochastic processes, in both discrete and continuous time, and provide sample complexity analysis extending existing theoretical guarantees for learning from non-sparse, i.i.d. data. In addressing continuous-time settings, we develop precise descriptions using covariance-type operators for the infinitesimal generator that aids in the sample complexity analysis. We empirically illustrate the efficacy of our sparse embedding approach through deterministic and stochastic nonlinear system examples.
Researcher Affiliation Academia 1Department of Electrical and Computer Engineering at the University of Illinois at Urbana-Champaign, Urbana, IL 61801, USA 2Department of Mechanical Engineering, Clemson University, Clemson SC, 29634, USA. Correspondence to: Boya Hou <boyahou2@illinois.edu>.
Pseudocode Yes Algorithm 1 Sample-based Sparse Learning of A Without Explicit Knowledge of SDE Coefficients
Open Source Code No The paper does not provide an explicit statement about releasing source code or a link to a code repository for the methodology described.
Open Datasets No The paper describes generating datasets for the experiments (e.g., 'we sample 100 trajectories to form D1 by first generating 100 uniformly distributed initial points from [−2, 2] × [−2, 2]', 'create a dataset D from 10 trajectories with 5000 evaluations each'), but it does not provide access information (like a URL, DOI, or specific citation) for a publicly available or open dataset.
Dataset Splits No The paper describes data generation and sub-sampling, but it does not specify explicit training, validation, or test dataset splits or mention cross-validation. For example, in Section 9.1: 'D2 is constructed from 50 uniformly distributed initial points with 100 evaluations along each and then sub-sampled with m = 20, s = 5 for each trajectory.'
Hardware Specification No The paper does not provide specific details about the hardware used for running the experiments (e.g., GPU/CPU models, memory specifications).
Software Dependencies No The paper mentions methods like 'Euler-Maruyama method (Higham, 2001)' but does not specify any software dependencies with version numbers (e.g., Python 3.x, PyTorch 1.x).
Experiment Setup Yes The paper provides detailed experimental setup for numerical experiments, including sampling intervals, number of trajectories, initial points, sub-sampling parameters, and kernel types. For example, in Section 9.1: 'To approximate A, we sample 100 trajectories to form D1 by first generating 100 uniformly distributed initial points from [−2, 2] × [−2, 2], then propagating them through the dynamics by evolving 1000 steps with sampling interval τ = 0.01 s. We then create a sub-sample with m = 200, and s = 5 along each trajectory. We utilize a Gaussian kernel κ(x1, x2) = exp(−‖x1 − x2‖2/(2σ2)), whose partial derivatives are included in Appendix N. Setting γ1 = 0.992, we get |Dγ1| = 1477.'