VaR-CPO optimizes VaR-constrained RL problems with conservative policy updates.
problem Optimizing VaR-constrained reinforcement learning problems.
method Combines Cantelli's inequality and trust-region framework for efficient and conservative optimization.
result Achieves zero constraint violations during training in feasible environments.
Novel methods generate diverse policies in reinforcement learning.
problem Generating diverse policies in reinforcement learning.
method Constrained optimization perspective, introducing new metrics, and novel policy generation methods.
result Improved novelty and performance of generated policies.
KL-constrained API shows optimization issues and improved with regularization.
problem Optimization issues in KL-constrained API algorithms.
method Comparison of KL divergence as a constraint vs. regularizer, empirical evaluation.
result KL-constrained API is not guaranteed to converge and incurs linear regret.
The paper optimizes policies constrained to Schur stabilizing controllers using a Newton-type algorithm.
problem Optimizing policies under linear constraints in control systems.
method Newton-type algorithm on a manifold of Schur stabilizing controllers with a Riemannian metric.
result Local convergence guarantees for the Newton-type algorithm without relying on exponential mapping or retractions.
In reinforcement learning, an agent attempts to learn high-performing behaviors through interacting with the environment, such behaviors are often quantified in the form of a reward function. However some aspects of behavior-such as ones which are deemed unsafe and to be avoided-are best captured through constraints. W…
Solving tasks in Reinforcement Learning is no easy feat. As the goal of the agent is to maximize the accumulated reward, it often learns to exploit loopholes and misspecifications in the reward signal resulting in unwanted behavior. While constraints may solve this issue, there is no closed form solution for general co…
Oracle-efficient algorithm for offline RL with partial data coverage.
problem Offline reinforcement learning with partial data coverage and constraints.
method PDOCRL, a primal-dual algorithm with decomposed linear-programming formulation.
result Near-optimal, near-feasible policy with \(\widetilde{\mathcal O}(ε^{-2})\) sample guarantee.
Offline RL tackles resource-constrained online deployment with improved policy transfer.
problem Training policies with limited online features using a rich offline dataset.
method Introduce a policy transfer algorithm that first trains a teacher agent with full offline features and then transfers knowledge to a student agent with limited online features.
result Consistent improvement in performance over baseline methods on resource-constrained datasets.
Canary optimizes VaR-constrained RL problems with a conservative bound using Cantelli's inequality.
problem Optimizing reinforcement learning policies under VaR constraints in dense cost regimes.
method Employing Cantelli's inequality to create a conservative and smooth bound on VaR constraints based on moments of cost returns. Extending trust-region framework for worst-case bounds on policy improvement and constraint violation.
result Canary reliably satisfies VaR constraints with fewest violations and earliest permanent satisfaction, while maintaining reward competitiveness.
In batch reinforcement learning (RL), one often constrains a learned policy to be close to the behavior (data-generating) policy, e.g., by constraining the learned action distribution to differ from the behavior policy by some maximum degree that is the same at each state. This can cause batch RL to be overly conservat…
Proposes Constrained Q-learning for reinforcement learning with constraints.
problem Optimizing multiple objectives while adhering to constraints in reinforcement learning.
method Directly restricts the action space in Q-update to learn optimal Q-function for constrained MDP.
result Improves safety and optimality in high-level decision making for autonomous driving.
PDCA algorithm learns policies for RL with constraints using a primal-dual approach.
problem Offline constrained reinforcement learning with general function approximation.
method Primal-Dual-Critic Algorithm (PDCA) using a primal-dual approach.
result PDCA finds a near saddle point of the Lagrangian, nearly optimal for constrained RL.
Paper proposes risk-averse reinforcement learning algorithms.
problem Managing model uncertainty in reinforcement learning.
method Entropic risk constrained policy gradient and actor-critic algorithms.
result Demonstrates usefulness of risk-averse algorithms on various domains.
We propose a policy improvement algorithm for Reinforcement Learning (RL) which is called Rerouted Behavior Improvement (RBI). RBI is designed to take into account the evaluation errors of the Q-function. Such errors are common in RL when learning the Q-value from finite past experience data. Greedy policies or even …
Algorithm optimizes constrained reinforcement learning with dual variables.
problem Minimizing convex functional subject to convex constraint in large state spaces.
method VPDPO algorithm using Lagrangian and Fenchel duality.
result Achieves sublinear regret and constraint violation, globally optimal policy.
KCRL learns stable policies for nonlinear systems with formal guarantees.
problem Lack of stabilization guarantees in RL methods for safety-critical systems.
method KCRL uses Krasovskii's Lyapunov functions as a stability constraint and a primal-dual approach to learn stabilizing policies.
result KCRL guarantees learning a stabilizing policy in a finite number of interactions.
A scalable framework optimizes multi-asset portfolios with constraints.
problem Optimizing multi-asset portfolios with inequality constraints.
method Integrates neural policies with Pontryagin's Maximum Principle, enforcing feasibility via log-barrier regularization.
result Recover KKT-optimal policies in high-dimensional problems without violating constraints.
Study optimal treatment assignment policies under strategic agent responses.
problem Learning optimal treatment policies with strategic agents complicates estimation.
method Dynamic model with threshold convergence to mean-field equilibrium, consistent estimator for policy gradient.
result Threshold for treatment assignment converges to mean-field equilibrium threshold under large but finite number of agents.
Safe RL for autonomous vehicles using PCPO with trust regions and parallel learners.
problem Unexplainable behaviours and lack of safety guarantees in RL for real vehicles.
method PCPO framework with trust regions and parallel learners.
result Safe learning confirmed for autonomous vehicles with fast convergence.
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 such a problem, we construct a sequence of surrogate convex constrained optimiza…
We study in this paper a class of constrained linear-quadratic (LQ) optimal control problem formulations for the scalar-state stochastic system with multiplicative noise, which has various applications, especially in the financial risk management. The linear constraint on both the control and state variables considered…
Off-policy deep reinforcement learning (RL) algorithms are incapable of learning solely from batch offline data without online interactions with the environment, due to the phenomenon known as \textit{extrapolation error}. This is often due to past data available in the replay buffer that may be quite different from th…
Many practical applications of reinforcement learning constrain agents to learn from a fixed batch of data which has already been gathered, without offering further possibility for data collection. In this paper, we demonstrate that due to errors introduced by extrapolation, standard off-policy deep reinforcement learn…
Develops neural network framework for risk-reward optimization problems.
problem Multi-period risk-reward optimization with constrained policies.
method Neural network framework with two coupled feedforward networks, parametrizing two-step policies.
result Empirical optimum converges to true optimal value as network capacity and training size increase.
This article examines arbitrage investment in a mispriced asset when the mispricing follows the Ornstein-Uhlenbeck process and a credit-constrained investor maximizes a generalization of the Kelly criterion. The optimal differentiable and threshold policies are derived. The optimal differentiable policy is linear with …
A novel approach for safe offline RL using latent safety constraints.
problem Balancing safety constraints and reward maximization in offline RL.
method Conditional Variational Autoencoders for latent safety modeling, Constrained Reward-Return Maximization.
result Our approach maintains safety compliance while optimizing rewards, outperforming existing methods.
We study continuous action reinforcement learning problems in which it is crucial that the agent interacts with the environment only through safe policies, i.e.,~policies that do not take the agent to undesirable situations. We formulate these problems as constrained Markov decision processes (CMDPs) and present safe p…
Batch Reinforcement Learning (Batch RL) consists in training a policy using trajectories collected with another policy, called the behavioural policy. Safe policy improvement (SPI) provides guarantees with high probability that the trained policy performs better than the behavioural policy, also called baseline in this…
This paper studies a class of continuous-time scalar-state stochastic Linear-Quadratic (LQ) optimal control problem with the linear control constraints. Applying the state separation theorem induced from its special structure, we develop the explicit solution for this class of problem. The revealed optimal control poli…
New algorithm reduces sample complexity for constrained MDPs.
problem Learning policies in constrained average-reward MDPs.
method Model-based algorithm for relaxed and strict feasibility settings.
result Achieves minimax-optimal bounds for constrained MDPs.
It is well known that the historical logs are used for evaluating and learning policies in interactive systems, e.g. recommendation, search, and online advertising. Since direct online policy learning usually harms user experiences, it is more crucial to apply off-policy learning in real-world applications instead. Tho…
New RL approach tackles constrained Markov decision processes.
problem Applying RL to physical systems with safety constraints.
method Formulated as a Constrained Markov Decision Process (CMDP), introduced a safe policy improvement method.
result Agent learns to maximize returns while satisfying constraints.
New method optimizes experiments under constraints.
problem Adapting BED to dynamic constraints in real-world tasks.
method Offline pre-training of an amortized policy and posterior network with online multi-step lookahead planning.
result Significantly more informative design sequences than existing methods.
State-constrained offline RL expands RL's learning scope.
problem Batch-constrained offline RL limits policies to seen actions.
method Introduces state-constrained offline RL focusing on state distribution.
result Policy can take high-quality out-of-distribution actions.
Dynamic Programming (DP) provides standard algorithms to solve Markov Decision Processes. However, these algorithms generally do not optimize a scalar objective function. In this paper, we draw connections between DP and (constrained) convex optimization. Specifically, we show clear links in the algorithmic structure b…
The paper studies consistency of surrogate loss procedures under constrained classifiers.
problem Consistency of surrogate loss approaches under constrained classifiers without correct specification.
method The paper develops theoretical results and hinge loss based procedures for a constrained classification problem.
result Hinge losses are the only surrogate losses that preserve consistency in second-best scenarios.
New algorithm for risk-sensitive reinforcement learning with natural policy gradients.
problem Risk-sensitive reinforcement learning with downside risk constraints.
method Introduce a new Bellman equation to estimate the lower partial moment of returns, use natural policy gradients, and extend Reward Constrained Policy Optimization.
result Sample-efficient estimation of partial moments and effective risk-sensitive control.
This work tackles risk-sensitive deep RL by optimizing policies with variance constraints.
problem Risk and aleatoric uncertainty in deep reinforcement learning.
method Lagrangian and Fenchel dualities to transform the problem into an unconstrained saddle-point policy optimization problem, and an actor-critic algorithm to iteratively update policy, Lagrange multiplier, and Fenchel dual variable.
result The proposed actor-critic algorithm finds a globally optimal policy at a sublinear rate.
Algorithm mitigates performance loss in constrained reinforcement learning with model misspecification.
problem Performance loss in reinforcement learning policies due to model misspecification in constrained control systems.
method Proposes an algorithm to handle constrained model misspecification in continuous control systems.
result Algorithm successfully mitigates performance loss in real-world reinforcement learning tasks.
Study examines sample complexity for RL with safety constraints.
problem Ensuring safety in RL for systems with constraints.
method Analyzes two RL approaches: generative and online.
result Sample complexity increased by a logarithmic factor in constraints.
New bounds for learning near-optimal policies in CMDPs with constraints.
problem Optimizing policies in CMDPs with constraints.
method Model-based algorithm addressing relaxed and strict feasibility.
result Near-optimal sample complexity bounds for CMDPs.
Develops new reinforcement learning methods for complex constrained decision-making problems.
problem Complex constrained decision-making problems with a continuum of constraints.
method Proposes semi-infinitely constrained Markov decision processes (SICMDPs) and two reinforcement learning algorithms: SI-CRL and SI-CPO.
result Demonstrates the effectiveness of SI-CRL and SI-CPO in solving complex sequential decision-making tasks.
The present work extends the randomized shortest-paths framework (RSP), interpolating between shortest-path and random-walk routing in a network, in three directions. First, it shows how to deal with equality constraints on a subset of transition probabilities and develops a generic algorithm for solving this constrain…
We address the problem of learning hierarchical deep neural network policies for reinforcement learning. In contrast to methods that explicitly restrict or cripple lower layers of a hierarchy to force them to use higher-level modulating signals, each layer in our framework is trained to directly solve the task, but acq…
Develops ODRPO to improve RL algorithms with better performance and stability.
problem RL algorithms converge to sub-optimal solutions due to limited policy representation.
method Integrates DRO approach to solve trust region constrained optimization problem without parameterizing policies.
result Achieves globally optimal policy update and higher sample efficiency.
We observe that several existing policy gradient methods (such as vanilla policy gradient, PPO, A2C) may suffer from overly large gradients when the current policy is close to deterministic (even in some very simple environments), leading to an unstable training process. To address this issue, we propose a new method, …
Policy gradient algorithms typically combine discounted future rewards with an estimated value function, to compute the direction and magnitude of parameter updates. However, for most Reinforcement Learning tasks, humans can provide additional insight to constrain the policy learning. We introduce a general method to i…
Neural Index Policy for multi-action bandits with heterogeneous budgets.
problem Real-world settings often involve multiple interventions with heterogeneous costs and constraints, breaking classical assumptions.
method Introduces a Neural Index Policy (NIP) that learns to assign budget-aware indices to arm-action pairs using a neural network and differentiable knapsack layer.
result Empirically achieves near-optimal performance while strictly enforcing heterogeneous budgets and scaling to hundreds of arms.