Saved in:
Bibliographic Details
Main Authors: Milosevic, Nikola, Scherf, Nico
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2509.01432
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908512430325760
author Milosevic, Nikola
Scherf, Nico
author_facet Milosevic, Nikola
Scherf, Nico
contents Reward maximization, safe exploration, and intrinsic motivation are often studied as separate objectives in reinforcement learning (RL). We present a unified geometric framework, that views these goals as instances of a single optimization problem on the space of achievable long-term behavior in an environment. Within this framework, classical methods such as policy mirror descent, natural policy gradient, and trust-region algorithms naturally generalize to nonlinear utilities and convex constraints. We illustrate how this perspective captures robustness, safety, exploration, and diversity objectives, and outline open challenges at the interface of geometry and deep RL.
format Preprint
id arxiv_https___arxiv_org_abs_2509_01432
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle The Geometry of Nonlinear Reinforcement Learning
Milosevic, Nikola
Scherf, Nico
Machine Learning
Reward maximization, safe exploration, and intrinsic motivation are often studied as separate objectives in reinforcement learning (RL). We present a unified geometric framework, that views these goals as instances of a single optimization problem on the space of achievable long-term behavior in an environment. Within this framework, classical methods such as policy mirror descent, natural policy gradient, and trust-region algorithms naturally generalize to nonlinear utilities and convex constraints. We illustrate how this perspective captures robustness, safety, exploration, and diversity objectives, and outline open challenges at the interface of geometry and deep RL.
title The Geometry of Nonlinear Reinforcement Learning
topic Machine Learning
url https://arxiv.org/abs/2509.01432