References & Citations
Mathematics > Optimization and Control
Title: A Measure Theoretical Approach to the Mean-field Maximum Principle for Training NeurODEs
(Submitted on 19 Jul 2021 (v1), last revised 8 Apr 2022 (this version, v2))
Abstract: In this paper we consider a measure-theoretical formulation of the training of NeurODEs in the form of a mean-field optimal control with $L^2$-regularization of the control. We derive first order optimality conditions for the NeurODE training problem in the form of a mean-field maximum principle, and show that it admits a unique control solution, which is Lipschitz continuous in time. As a consequence of this uniqueness property, the mean-field maximum principle also provides a strong quantitative generalization error for finite sample approximations. Our derivation of the mean-field maximum principle is much simpler than the ones currently available in the literature for mean-field optimal control problems, and is based on a generalized Lagrange multiplier theorem on convex sets of spaces of measures. The latter is also new, and can be considered as a result of independent interest.
Submission history
From: Cristina Cipriani [view email][v1] Mon, 19 Jul 2021 09:31:33 GMT (2534kb,D)
[v2] Fri, 8 Apr 2022 15:22:04 GMT (11657kb,D)
Link back to: arXiv, form interface, contact.