arXiv research

Novel framework for policy optimization with general parameterization and linear convergence.

We propose a new sample-efficient methodology, called Supervised Policy Update (SPU), for deep reinforcement learning. Starting with data generated by the current policy, SPU formulates and solves a constrained optimization problem in the non-parameterized proximal policy space. Using supervised regression, it then con…

We introduce a method for constructing skills capable of solving tasks drawn from a distribution of parameterized reinforcement learning problems. The method draws example tasks from a distribution of interest and uses the corresponding learned policies to estimate the topology of the lower-dimensional piecewise-smooth…

Policy gradient converges linearly with Hadamard parameterization in tabular settings.

Revisits PPO design choices, exposing failure modes and proposing alternatives.

We explore Deep Reinforcement Learning in a parameterized action space. Specifically, we investigate how to achieve sample-efficient end-to-end training in these tasks. We propose a new compact architecture for the tasks where the parameter policy is conditioned on the output of the discrete action policy. We also prop…

Policy gradient based reinforcement learning algorithms coupled with neural networks have shown success in learning complex policies in the model free continuous action space control setting. However, explicitly parameterized policies are limited by the scope of the chosen parametric probability distribution. We show t…

Double descent found in DRL, improving generalization with model capacity.

In this paper we propose a hybrid architecture of actor-critic algorithms for reinforcement learning in parameterized action space, which consists of multiple parallel sub-actor networks to decompose the structured action space into simpler action spaces along with a critic network to guide the training of all sub-acto…

Branch and Bound (B&B) is the exact tree search method typically used to solve Mixed-Integer Linear Programming problems (MILPs). Learning branching policies for MILP has become an active research area, with most works proposing to imitate the strong branching rule and specialize it to distinct classes of problems. We …

We compare the model-free reinforcement learning with the model-based approaches through the lens of the expressive power of neural networks for policies, $Q$-functions, and dynamics. We show, theoretically and empirically, that even for one-dimensional continuous state space, there are many MDPs whose optimal $Q$-func…

A new method for learning policies from demonstrations without reinforcement.

We present a novel solution to the problem of simulation-to-real transfer, which builds on recent advances in robot skill decomposition. Rather than focusing on minimizing the simulation-reality gap, we learn a set of diverse policies that are parameterized in a way that makes them easily reusable. This diversity and p…

The recently proposed option-critic architecture Bacon et al. provide a stochastic policy gradient approach to hierarchical reinforcement learning. Specifically, they provide a way to estimate the gradient of the expected discounted return with respect to parameters that define a finite number of temporally extended ac…

Study improves policy search in continuous control by using heavy-tailed distributions.

This paper identifies drift Lipschitz budget K as key to diffusion policy expressivity and statistical trade-offs.

Survey explores geometric aspects of policy optimization in control systems.

Due to the high variance of policy gradients, on-policy optimization algorithms are plagued with low sample efficiency. In this work, we propose Augment-Reinforce-Merge (ARM) policy gradient estimator as an unbiased low-variance alternative to previous baseline estimators on tasks with binary action space, inspired by …

Learning optimal resource allocation policies in wireless systems can be effectively achieved by formulating finite dimensional constrained programs which depend on system configuration, as well as the adopted learning parameterization. The interest here is in cases where system models are unavailable, prompting method…

Inspired by the seminal work on Stein Variational Inference and Stein Variational Policy Gradient, we derived a method to generate samples from the posterior variational parameter distribution by \textit{explicitly} minimizing the KL divergence to match the target distribution in an amortize fashion. Consequently, we a…

Optimizes wireless network resource management with state-augmented policies.

Continuous control imitation learning fails if expert actions are smooth.

Agent57 beats human average on all Atari games.

Entropy regularization improves policy optimization in reinforcement learning.

New IRL algorithm identifies optimal reward and policy from expert demonstrations.

Develops ODRPO to improve RL algorithms with better performance and stability.

Trust-region methods have yielded state-of-the-art results in policy search. A common approach is to use KL-divergence to bound the region of trust resulting in a natural gradient policy update. We show that the natural gradient and trust region optimization are equivalent if we use the natural parameterization of a st…

The most data-efficient algorithms for reinforcement learning in robotics are model-based policy search algorithms, which alternate between learning a dynamical model of the robot and optimizing a policy to maximize the expected return given the model and its uncertainties. Among the few proposed approaches, the recent…

The paper tackles batch policy learning in Markov Decision Processes, focusing on average reward maximization.

Graph neural networks optimize radio resource management policies for wireless networks.

Exploration policies in Bayesian bandits maximize the average reward over problem instances drawn from some distribution $\mathcal{P}$. In this work, we learn such policies for an unknown distribution $\mathcal{P}$ using samples from $\mathcal{P}$. Our approach is a form of meta-learning and exploits properties of $\ma…

Experience replay (ER) is a fundamental component of off-policy deep reinforcement learning (RL). ER recalls experiences from past iterations to compute gradient estimates for the current policy, increasing data-efficiency. However, the accuracy of such updates may deteriorate when the policy diverges from past behavio…

Policy gradients methods apply to complex, poorly understood, control problems by performing stochastic gradient descent over a parameterized class of polices. Unfortunately, even for simple control problems solvable by standard dynamic programming techniques, policy gradient algorithms face non-convex optimization pro…

This work improves policy-based training by proposing an evaluation balance objective for GFlowNets.

Generative Adversarial Imitation Learning (GAIL) is a powerful and practical approach for learning sequential decision-making policies. Different from Reinforcement Learning (RL), GAIL takes advantage of demonstration data by experts (e.g., human), and learns both the policy and reward function of the unknown environme…

This paper introduces the QMDP-net, a neural network architecture for planning under partial observability. The QMDP-net combines the strengths of model-free learning and model-based planning. It is a recurrent policy network, but it represents a policy for a parameterized set of tasks by connecting a model with a plan…

We present GradientDICE for estimating the density ratio between the state distribution of the target policy and the sampling distribution in off-policy reinforcement learning. GradientDICE fixes several problems of GenDICE (Zhang et al., 2020), the state-of-the-art for estimating such density ratios. Namely, the optim…

Unified approach to Merton's portfolio problem using Pontryagin's principles.

We study Bayesian optimal control of a general class of smoothly parameterized Markov decision problems. Since computing the optimal control is computationally expensive, we design an algorithm that trades off performance for computational efficiency. The algorithm is a lazy posterior sampling method that maintains a d…

Optimizes insurance pricing by accounting for policyholders' price sensitivity.

Paper proposes FPG algorithm for unbiased off-policy PG estimation.

A method for accurate pricing of multidimensional derivatives under uncertain volatility.

A large body of animation research focuses on optimization of movement control, either as action sequences or policy parameters. However, as closed-form expressions of the objective functions are often not available, our understanding of the optimization problems is limited. Building on recent work on analyzing neural …

Entropy-regularized NPG methods converge linearly in discounted MDPs.

Improved exploration in cooperative multi-agent reinforcement learning.

Policy gradient methods with actor-critic schemes demonstrate tremendous empirical successes, especially when the actors and critics are parameterized by neural networks. However, it remains less clear whether such "neural" policy gradient methods converge to globally optimal policies and whether they even converge at …

We propose the Insertion-Deletion Transformer, a novel transformer-based neural architecture and training method for sequence generation. The model consists of two phases that are executed iteratively, 1) an insertion phase and 2) a deletion phase. The insertion phase parameterizes a distribution of insertions on the c…

Policy gradient methods are among the most effective methods in challenging reinforcement learning problems with large state and/or action spaces. However, little is known about even their most basic theoretical convergence properties, including: if and how fast they converge to a globally optimal solution or how they …