Approximately Gaussian Replicator Flows: Nonconvex Optimization as a Nash-Convergent Evolutionary Game

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Anderson, Brendon G., Pfrommer, Samuel, Sojoudi, Somayeh
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866929493803794432
author Anderson, Brendon G.
Pfrommer, Samuel
Sojoudi, Somayeh
author_facet Anderson, Brendon G.
Pfrommer, Samuel
Sojoudi, Somayeh
contents This work leverages tools from evolutionary game theory to solve unconstrained nonconvex optimization problems. Specifically, we lift such a problem to an optimization over probability measures, whose minimizers exactly correspond to the Nash equilibria of a particular population game. To algorithmically solve for such Nash equilibria, we introduce approximately Gaussian replicator flows (AGRFs) as a tractable alternative to simulating the corresponding infinite-dimensional replicator dynamics. Our proposed AGRF dynamics can be integrated using off-the-shelf ODE solvers when considering objectives with closed-form integrals against a Gaussian measure. We theoretically analyze AGRF dynamics by explicitly characterizing their trajectories and stability on quadratic objective functions, in addition to analyzing their descent properties. Our methods are supported by illustrative experiments on a range of canonical nonconvex optimization benchmark functions.
format Preprint
id arxiv_https___arxiv_org_abs_2406_19529
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Approximately Gaussian Replicator Flows: Nonconvex Optimization as a Nash-Convergent Evolutionary Game
Anderson, Brendon G.
Pfrommer, Samuel
Sojoudi, Somayeh
Optimization and Control
This work leverages tools from evolutionary game theory to solve unconstrained nonconvex optimization problems. Specifically, we lift such a problem to an optimization over probability measures, whose minimizers exactly correspond to the Nash equilibria of a particular population game. To algorithmically solve for such Nash equilibria, we introduce approximately Gaussian replicator flows (AGRFs) as a tractable alternative to simulating the corresponding infinite-dimensional replicator dynamics. Our proposed AGRF dynamics can be integrated using off-the-shelf ODE solvers when considering objectives with closed-form integrals against a Gaussian measure. We theoretically analyze AGRF dynamics by explicitly characterizing their trajectories and stability on quadratic objective functions, in addition to analyzing their descent properties. Our methods are supported by illustrative experiments on a range of canonical nonconvex optimization benchmark functions.
title Approximately Gaussian Replicator Flows: Nonconvex Optimization as a Nash-Convergent Evolutionary Game
topic Optimization and Control
url https://arxiv.org/abs/2406.19529