Nonconvex Regularization for Feature Selection in Reinforcement Learning

• URL: http://arxiv.org/abs/2509.15652v1
• Date: Fri, 19 Sep 2025 06:21:20 GMT
• Title: Nonconvex Regularization for Feature Selection in Reinforcement Learning
• Authors: Kyohei Suzuki, Konstantinos Slavakis,
• Abstract summary: This work proposes an efficient batch algorithm for feature selection in reinforcement learning (RL) with theoretical convergence guarantees.<n> Numerical experiments demonstrate that the proposed approach substantially outperforms state-selection scenarios.
• Score: 7.408148824204063
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: This work proposes an efficient batch algorithm for feature selection in reinforcement learning (RL) with theoretical convergence guarantees. To mitigate the estimation bias inherent in conventional regularization schemes, the first contribution extends policy evaluation within the classical least-squares temporal-difference (LSTD) framework by formulating a Bellman-residual objective regularized with the sparsity-inducing, nonconvex projected minimax concave (PMC) penalty. Owing to the weak convexity of the PMC penalty, this formulation can be interpreted as a special instance of a general nonmonotone-inclusion problem. The second contribution establishes novel convergence conditions for the forward-reflected-backward splitting (FRBS) algorithm to solve this class of problems. Numerical experiments on benchmark datasets demonstrate that the proposed approach substantially outperforms state-of-the-art feature-selection methods, particularly in scenarios with many noisy features.

Related papers

• Principled Algorithms for Optimizing Generalized Metrics in Binary Classification [53.604375124674796]
  We introduce principled algorithms for optimizing generalized metrics, supported by $H$-consistency and finite-sample generalization bounds.<n>Our approach reformulates metric optimization as a generalized cost-sensitive learning problem.<n>We develop new algorithms, METRO, with strong theoretical performance guarantees.
  arXiv  Detail & Related papers  (2025-12-29T01:33:42Z)
• Latent Chain-of-Thought for Visual Reasoning [53.541579327424046]
  Chain-of-thought (CoT) reasoning is critical for improving the interpretability and reliability of Large Vision-Language Models (LVLMs)<n>We reformulate reasoning in LVLMs as posterior inference and propose a scalable training algorithm based on amortized variational inference.<n>We empirically demonstrate that the proposed method enhances the state-of-the-art LVLMs on seven reasoning benchmarks.
  arXiv  Detail & Related papers  (2025-10-27T23:10:06Z)
• Risk-Averse Constrained Reinforcement Learning with Optimized Certainty Equivalents [29.698100324454362]
  Constrained optimization provides a common framework for dealing with conflicting objectives in reinforcement learning (RL)<n>We propose a framework for risk-aware constrained RL, which exhibits per-stage properties jointly in reward values and time using optimized certainty equivalents (OCEs)<n>Our framework ensures an exact equivalent to the original constrained problem within a parameterized strong Lagrangian duality framework under appropriate constraint qualifications.
  arXiv  Detail & Related papers  (2025-10-23T04:33:32Z)
• Navigating Sparsities in High-Dimensional Linear Contextual Bandits [7.796930035720785]
  High-dimensional linear contextual bandit problems remain a significant challenge due to the curse of dimensionality.<n>A powerful pointwise estimator is introduced in this work that adaptively navigates both kinds of sparsity.<n>Based on this pointwise estimator, a novel algorithm, termed HOPE, is proposed.
  arXiv  Detail & Related papers  (2025-10-09T16:47:14Z)
• Risk-sensitive Reinforcement Learning Based on Convex Scoring Functions [8.758206783988404]
  We propose a reinforcement learning framework under a broad class of risk objectives, characterized by convex scoring functions.<n>This class covers many common risk measures, such as variance, Expected Shortfall, entropic Value-at-Risk, and mean-risk utility.<n>We validate our approach in simulation experiments with a financial application in statistical arbitrage trading, demonstrating the effectiveness of the algorithm.
  arXiv  Detail & Related papers  (2025-05-07T16:31:42Z)
• Single-loop Algorithms for Stochastic Non-convex Optimization with Weakly-Convex Constraints [49.76332265680669]
  This paper examines a crucial subset of problems where both the objective and constraint functions are weakly convex.<n>Existing methods often face limitations, including slow convergence rates or reliance on double-loop designs.<n>We introduce a novel single-loop penalty-based algorithm to overcome these challenges.
  arXiv  Detail & Related papers  (2025-04-21T17:15:48Z)
• Effectively Leveraging Momentum Terms in Stochastic Line Search Frameworks for Fast Optimization of Finite-Sum Problems [1.5665889790660235]
  We explore the relationship between recent line search approaches for deep optimization and momentum directions.<n>We introduce an algorithmic framework that exploits a mix of data persistency, persistient type rules for definition of the momentum parameter and line searches.<n>The resulting algorithm provably possesses convergence under suitable assumptions.
  arXiv  Detail & Related papers  (2024-11-11T16:26:33Z)
• Constrained Sampling with Primal-Dual Langevin Monte Carlo [15.634831573546041]
  This work considers the problem of sampling from a probability distribution known up to a normalization constant.<n>It satisfies a set of statistical constraints specified by the expected values of general nonlinear functions.<n>We put forward a discrete-time primal-dual Langevin Monte Carlo algorithm (PD-LMC) that simultaneously constrains the target distribution and samples from it.
  arXiv  Detail & Related papers  (2024-11-01T13:26:13Z)
• Generalization Bounds of Surrogate Policies for Combinatorial Optimization Problems [53.03951222945921]
  We analyze smoothed (perturbed) policies, adding controlled random perturbations to the direction used by the linear oracle.<n>Our main contribution is a generalization bound that decomposes the excess risk into perturbation bias, statistical estimation error, and optimization error.<n>We illustrate the scope of the results on applications such as vehicle scheduling, highlighting how smoothing enables both tractable training and controlled generalization.
  arXiv  Detail & Related papers  (2024-07-24T12:00:30Z)
• Likelihood Ratio Confidence Sets for Sequential Decision Making [51.66638486226482]
  We revisit the likelihood-based inference principle and propose to use likelihood ratios to construct valid confidence sequences. Our method is especially suitable for problems with well-specified likelihoods. We show how to provably choose the best sequence of estimators and shed light on connections to online convex optimization.
  arXiv  Detail & Related papers  (2023-11-08T00:10:21Z)
• Online Non-convex Optimization with Long-term Non-convex Constraints [7.68925638410622]
  A novel Follow-the-Perturbed-Leader type algorithm is proposed and analyzed for solving general long-term constrained optimization problems in an online manner.<n>The proposed is applied to tackle a long-term constrained source identification problem, validate the theoretical results and exhibit superior performance compared to existing methods.
  arXiv  Detail & Related papers  (2023-11-04T15:08:36Z)
• Predictor-Rejector Multi-Class Abstention: Theoretical Analysis and Algorithms [30.389055604165222]
  We study the key framework of learning with abstention in the multi-class classification setting. In this setting, the learner can choose to abstain from making a prediction with some pre-defined cost. We introduce several new families of surrogate losses for which we prove strong non-asymptotic and hypothesis set-specific consistency guarantees.
  arXiv  Detail & Related papers  (2023-10-23T10:16:27Z)
• GEC: A Unified Framework for Interactive Decision Making in MDP, POMDP, and Beyond [101.5329678997916]
  We study sample efficient reinforcement learning (RL) under the general framework of interactive decision making. We propose a novel complexity measure, generalized eluder coefficient (GEC), which characterizes the fundamental tradeoff between exploration and exploitation. We show that RL problems with low GEC form a remarkably rich class, which subsumes low Bellman eluder dimension problems, bilinear class, low witness rank problems, PO-bilinear class, and generalized regular PSR.
  arXiv  Detail & Related papers  (2022-11-03T16:42:40Z)
• Finite Sample Analysis of Minimax Offline Reinforcement Learning: Completeness, Fast Rates and First-Order Efficiency [83.02999769628593]
  We offer a theoretical characterization of off-policy evaluation (OPE) in reinforcement learning. We show that the minimax approach enables us to achieve a fast rate of convergence for weights and quality functions. We present the first finite-sample result with first-order efficiency in non-tabular environments.
  arXiv  Detail & Related papers  (2021-02-05T03:20:39Z)
• Selective Classification via One-Sided Prediction [54.05407231648068]
  One-sided prediction (OSP) based relaxation yields an SC scheme that attains near-optimal coverage in the practically relevant high target accuracy regime. We theoretically derive bounds generalization for SC and OSP, and empirically we show that our scheme strongly outperforms state of the art methods in coverage at small error levels.
  arXiv  Detail & Related papers  (2020-10-15T16:14:27Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.