Year: 2025
Type: article
Abstract: . Kakade’s natural policy gradient method has been studied extensively in recent years, showing linear convergence with and without regularization. We study another natural gradient method based on the Fisher information matrix of the state-action distributions which has received little attention from the theoretical side. Here, the state-action distributions follow the Fisher–Rao gradient flow inside the state-action polytope with respect to a linear potential. Therefore, we study Fisher–Rao gradient flows of linear programs more generally and show linear convergence with a rate that depends on the geometry of the linear program. Equivalently, this yields an estimate on the error induced by entropic regularization of the linear program which improves existing results. We extend these results and show sublinear convergence for perturbed Fisher–Rao gradient flows and natural gradient flows up to an approximation error. In particular, these general results cover the case of state-action natural policy gradients.
Source: SIAM Journal on Optimization
DOI: 10.1137/24m1653422
Authors: Johannes Müller, Semih Çayci, Guido Montúfar
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
Funders: National Science Foundation, International Max Planck Research School for Environmental, Cellular and Molecular Microbiology, Deutscher Akademischer Austauschdienst, Deutsche Forschungsgemeinschaft, Bundesministerium für Bildung und Forschung, Evangelisches Studienwerk Villigst, Max-Planck-Institut für Mathematik in den Naturwissenschaften, European Research Council
Awards: Kombinatorische und implizite Methoden für Deep Learning, G442688124, G4712284355, G4771048887, Collaborative Research: RI: Medium: MoDL: Occams Razor in Deep and Physical Learning, G6098565189, G6337610878, G6671297155, SFB 1481: Sparsity und singuläre Strukturen, CAREER: Neural Networks in the Practical Regime
OpenAlex ID: W4410413636
