English
Related papers

Related papers: Rethinking the Global Convergence of Softmax Polic…

200 papers

In this work, we introduce a stochastic maximum principle (SMP) approach for solving the reinforcement learning problem with the assumption that the unknowns in the environment can be parameterized based on physics knowledge. For the…

Optimization and Control · Mathematics 2023-06-14 Richard Archibald , Feng Bao , Jiongmin Yong

The infinite horizon setting is widely adopted for problems of reinforcement learning (RL). These invariably result in stationary policies that are optimal. In many situations, finite horizon control problems are of interest and for such…

Machine Learning · Computer Science 2025-03-21 Soumyajit Guin , Shalabh Bhatnagar

We study a reinforcement learning setting, where the state transition function is a convex combination of a stochastic continuous function and a deterministic function. Such a setting generalizes the widely-studied stochastic state…

Machine Learning · Computer Science 2018-10-03 Qingpeng Cai , Ling Pan , Pingzhong Tang

In recent years, reinforcement learning (RL) systems with general goals beyond a cumulative sum of rewards have gained traction, such as in constrained problems, exploration, and acting upon prior experiences. In this paper, we consider…

Machine Learning · Computer Science 2020-07-07 Junyu Zhang , Alec Koppel , Amrit Singh Bedi , Csaba Szepesvari , Mengdi Wang

Policy gradient methods have become a staple of any single-agent reinforcement learning toolbox, due to their combination of desirable properties: iterate convergence, efficient use of stochastic trajectory feedback, and theoretically-sound…

Computer Science and Game Theory · Computer Science 2025-07-10 Mingyang Liu , Gabriele Farina , Asuman Ozdaglar

Reward-Weighted Regression (RWR) belongs to a family of widely known iterative Reinforcement Learning algorithms based on the Expectation-Maximization framework. In this family, learning at each iteration consists of sampling a batch of…

In reinforcement learning, the softmax parametrization is the standard approach for policies over discrete action spaces. However, it fails to capture the order relationship between actions. Motivated by a real-world industrial problem, we…

Machine Learning · Computer Science 2025-06-24 Simón Weinberger , Jairo Cugliari

We study the safe reinforcement learning problem with nonlinear function approximation, where policy optimization is formulated as a constrained optimization problem with both the objective and the constraint being nonconvex functions. For…

Machine Learning · Computer Science 2019-10-29 Ming Yu , Zhuoran Yang , Mladen Kolar , Zhaoran Wang

Policy gradient methods are notorious for having a large variance and high sample complexity. To mitigate this, we introduce SoftTreeMax -- a generalization of softmax that employs planning. In SoftTreeMax, we extend the traditional logits…

Machine Learning · Computer Science 2025-05-27 Gal Dalal , Assaf Hallak , Gugan Thoppe , Shie Mannor , Gal Chechik

Policy gradient algorithms have been widely applied to Markov decision processes and reinforcement learning problems in recent years. Regularization with various entropy functions is often used to encourage exploration and improve…

Machine Learning · Computer Science 2023-06-09 Haoya Li , Samarth Gupta , Hsiangfu Yu , Lexing Ying , Inderjit Dhillon

We establish a new connection between value and policy based reinforcement learning (RL) based on a relationship between softmax temporal value consistency and policy optimality under entropy regularization. Specifically, we show that…

Artificial Intelligence · Computer Science 2017-11-27 Ofir Nachum , Mohammad Norouzi , Kelvin Xu , Dale Schuurmans

We consider policy gradient methods for stochastic optimal control problem in continuous time. In particular, we analyze the gradient flow for the control, viewed as a continuous time limit of the policy gradient method. We prove the global…

Optimization and Control · Mathematics 2025-04-15 Mo Zhou , Jianfeng Lu

Policy gradient methods, which have been extensively studied in the last decade, offer an effective and efficient framework for reinforcement learning problems. However, their performances can often be unsatisfactory, suffering from…

Machine Learning · Computer Science 2026-01-27 Shihab Ahmed , El Houcine Bergou , Aritra Dutta , Yue Wang

Estimating a policy that maps states to actions is a central problem in reinforcement learning. Traditionally, policies are inferred from the so called value functions (VFs), but exact VF computation suffers from the curse of…

Machine Learning · Computer Science 2024-05-29 Sergio Rozada , Antonio G. Marques

Policy-based algorithms are among the most widely adopted techniques in model-free RL, thanks to their strong theoretical groundings and good properties in continuous action spaces. Unfortunately, these methods require precise and…

Machine Learning · Computer Science 2023-06-14 Luca Sabbioni , Francesco Corda , Marcello Restelli

We show that the \emph{stochastic gradient} bandit algorithm converges to a \emph{globally optimal} policy at an $O(1/t)$ rate, even with a \emph{constant} step size. Remarkably, global convergence of the stochastic gradient bandit…

Machine Learning · Computer Science 2024-02-28 Jincheng Mei , Zixin Zhong , Bo Dai , Alekh Agarwal , Csaba Szepesvari , Dale Schuurmans

Policy gradient (PG) methods are a class of effective reinforcement learning algorithms, particularly when dealing with continuous control problems. They rely on fresh on-policy data, making them sample-inefficient and requiring…

Machine Learning · Computer Science 2026-02-03 Alessandro Montenegro , Federico Mansutti , Marco Mussi , Matteo Papini , Alberto Maria Metelli

Policy gradient methods are known to be highly sensitive to the choice of policy parameterization. In particular, the widely used softmax parameterization can induce ill-conditioned optimization landscapes and lead to exponentially slow…

Machine Learning · Computer Science 2026-04-02 Safwan Labbi , Daniil Tiapkin , Paul Mangold , Eric Moulines

Goal-Conditioned Reinforcement Learning (RL) problems often have access to sparse rewards where the agent receives a reward signal only when it has achieved the goal, making policy optimization a difficult problem. Several works augment…

Machine Learning · Computer Science 2023-10-11 Siddhant Agarwal , Ishan Durugkar , Peter Stone , Amy Zhang

We present a methodology to deploy the stochastic policy gradient method, using actor-critic techniques, when the optimal policy is approximated using a parametric optimization problem, allowing one to enforce safety via hard constraints.…

Systems and Control · Electrical Eng. & Systems 2024-09-23 Sebastien Gros , Mario Zanon