This paper identifies drift Lipschitz budget K as key to diffusion policy expressivity and statistical trade-offs.
problem Understanding and maximizing the expressivity of diffusion policies while managing statistical limitations.
method Identifying drift Lipschitz budget K as central, quantifying expressivity and statistical behavior, proving lower bounds, and providing practical implementation guidelines.
result Balancing expressivity and statistical complexity yields a finite-sample performance gap, with rates depending on sample size and drift type.
Off-policy reinforcement learning (RL) is concerned with learning a rewarding policy by executing another policy that gathers samples of experience. While the former policy (i.e. target policy) is rewarding but in-expressive (in most cases, deterministic), doing well in the latter task, in contrast, requires an express…
Diffusion-QL uses diffusion models to improve offline RL performance.
problem Offline RL struggles with function approximation errors on out-of-distribution actions.
method Diffusion-QL represents the policy as a conditional diffusion model and optimizes action-values.
result Diffusion-QL achieves state-of-the-art performance on D4RL benchmark tasks.
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…
We introduce Implicit Policy, a general class of expressive policies that can flexibly represent complex action distributions in reinforcement learning, with efficient algorithms to compute entropy regularized policy gradients. We empirically show that, despite its simplicity in implementation, entropy regularization c…
QAM uses adjoint matching to optimize continuous-action RL policies efficiently.
problem Efficient optimization of expressive diffusion or flow-matching policies with respect to a Q-function.
method QAM leverages adjoint matching to bypass the numerical instability of backpropagation through multi-step denoising processes.
result QAM consistently outperforms prior approaches on hard, sparse reward tasks in offline and offline-to-online RL.
This work explains why online imitation learning improves faster than theory predicts.
problem Online imitation learning's empirical policy improvement speed exceeds theoretical predictions.
method The authors analyze online imitation learning with a convex, smooth, and non-negative loss function, proving policy improvement in expectation and high probability.
result Adopting a sufficiently expressive policy class in online IL increases both policy improvement speed and performance bias.
Improved exploration in cooperative multi-agent reinforcement learning.
problem Limited expressiveness of Gaussian policies in DecSPG hinders effective exploration.
method Proposes decentralized diffusion policy learning (DDPL) with denoising diffusion probabilistic models.
result Consistently improved performance on various MARL benchmarks.
Discovering the underlying mathematical expressions describing a dataset is a core challenge for artificial intelligence. This is the problem of symbolic regression. Despite recent advances in training neural networks to solve complex tasks, deep learning approaches to symbolic regression are underexplored. …
The paper explores learning good policies from past data in large state spaces.
problem Learning good policies from historical data in large state spaces.
method Introduces expressivity assumptions and data coverage for function approximation and algorithmic design.
result A variety of algorithms and their guarantees are presented based on assumptions and desired complexity.
We investigate statistical uncertainty quantification for reinforcement learning (RL) and its implications in exploration policy. Despite ever-growing literature on RL applications, fundamental questions about inference and error quantification, such as large-sample behaviors, appear to remain quite open. In this paper…
New algorithm stabilizes RL policy learning through divergence regularization.
problem Stabilize policy learning and improve performance in RL.
method Proximity term constraining discounted state-action visitation distributions to be close to each other.
result Proposed algorithm improves stability and final performance in RL tasks.
We develop a normative framework for hierarchical model-based policy optimization based on applying second-order methods in the space of all possible state-action paths. The resulting natural path gradient performs policy updates in a manner which is sensitive to the long-range correlational structure of the induced st…
A nonparametric approach for policy learning for POMDPs is proposed. The approach represents distributions over the states, observations, and actions as embeddings in feature spaces, which are reproducing kernel Hilbert spaces. Distributions over states given the observations are obtained by applying the kernel Bayes' …
Accelerates policy optimization in RL with optimistic and adaptive updates.
problem Improving policy optimization methods in reinforcement learning.
method Integrates foresight into policy improvement step via optimistic and adaptive updates.
result Designs an optimistic policy gradient algorithm, adaptive via meta-gradient learning.
Exploration policies in Bayesian bandits maximize the average reward over problem instances drawn from some distribution P. In this work, we learn such policies for an unknown distribution P using samples from P. Our approach is a form of meta-learning and exploits properties of $\ma…
Deep Reinforcement Learning (DRL) algorithms for continuous action spaces are known to be brittle toward hyperparameters as well as \cut{being}sample inefficient. Soft Actor Critic (SAC) proposes an off-policy deep actor critic algorithm within the maximum entropy RL framework which offers greater stability and empiric…
This paper considers the problem of optimal liquidation of a position in a risky security in a financial market, where price evolution are risky and trades have an impact on price as well as uncertainty in the filling orders. The problem is formulated as a continuous time stochastic optimal control problem aiming at ma…
ESAN improves graph neural networks by processing subgraphs.
problem Limitations of MPNNs in graph isomorphism.
method ESAN represents graphs as sets of subgraphs and processes them with equivariant architectures.
result ESAN increases the expressive power of GNNs and more expressive architectures.
Reinforcement learning (RL) algorithms still suffer from high sample complexity despite outstanding recent successes. The need for intensive interactions with the environment is especially observed in many widely popular policy gradient algorithms that perform updates using on-policy samples. The price of such ineffici…
A Relational Markov Decision Process (RMDP) is a first-order representation to express all instances of a single probabilistic planning domain with possibly unbounded number of objects. Early work in RMDPs outputs generalized (instance-independent) first-order policies or value functions as a means to solve all instanc…
We propose a new objective function for finite-horizon episodic Markov decision processes that better captures Bellman's principle of optimality, and provide an expression for the gradient of the objective.
The paper tackles personalized policy learning from diverse data sources in a federated setting.
problem Learning personalized decision policies from observational bandit feedback across multiple heterogeneous data sources.
method Introduces a novel regret analysis for distinguishing global and local regret, and presents a federated policy learning algorithm using local policies trained with doubly robust offline policy evaluation strategies.
result Establishes finite-sample upper bounds on global and local regret, characterizing them by source heterogeneity and distribution shift.
In this work, we take a fresh look at some old and new algorithms for off-policy, return-based reinforcement learning. Expressing these in a common form, we derive a novel algorithm, Retrace(λ), with three desired properties: (1) it has low variance; (2) it safely uses samples collected from any behaviour policy, wha…
In recent years, state-of-the-art game-playing agents often involve policies that are trained in self-playing processes where Monte Carlo tree search (MCTS) algorithms and trained policies iteratively improve each other. The strongest results have been obtained when policies are trained to mimic the search behaviour of…
In this paper, we introduce a novel form of value function, Q(s,s′), that expresses the utility of transitioning from a state s to a neighboring state s′ and then acting optimally thereafter. In order to derive an optimal policy, we develop a forward dynamics model that learns to make next-state predictions that…
In this paper we study a class of insurance products where the policy holder has the option to insure k of its annual Operational Risk losses in a horizon of T years. This involves a choice of k out of T years in which to apply the insurance policy coverage by making claims against losses in the given year. The…
We present Multitask Soft Option Learning(MSOL), a hierarchical multitask framework based on Planning as Inference. MSOL extends the concept of options, using separate variational posteriors for each task, regularized by a shared prior. This ''soft'' version of options avoids several instabilities during training in a …
Extends double linear policy with time-varying weights and proves robust positive expectation.
problem Ensuring robustness in policy optimization with time-varying parameters.
method Employed a novel elementary symmetric polynomials characterization approach to prove robust positive expectation (RPE). Derived explicit expressions for expected cumulative gain-loss and variance.
result Proved the robust positive expectation property holds for the extended double linear policy.
Search-based methods for hard combinatorial optimization are often guided by heuristics. Tuning heuristics in various conditions and situations is often time-consuming. In this paper, we propose NeuRewriter that learns a policy to pick heuristics and rewrite the local components of the current solution to iteratively i…
Develops DPG methods for continuous-time RL with deterministic policies.
problem High variance and slow convergence in stochastic policy RL methods.
method Derives continuous-time policy gradient formula and proposes CT-DDPG algorithm.
result CT-DDPG achieves superior stability and faster convergence in continuous-time RL.
This paper presents the concept of an adaptive safe padding that forces Reinforcement Learning (RL) to synthesise optimal control policies while ensuring safety during the learning process. Policies are synthesised to satisfy a goal, expressed as a temporal logic formula, with maximal probability. Enforcing the RL agen…
In the hypothesis of rare loss events, the general expression of the policy value has been determined as a functional of the "expected frequency / loss severity" function and of the retention function. Exponential disutility has been chosen after mathematical characterization of some of its economical aspects, where fu…
We reformulate the option framework as two parallel augmented MDPs. Under this novel formulation, all policy optimization algorithms can be used off the shelf to learn intra-option policies, option termination conditions, and a master policy over options. We apply an actor-critic algorithm on each augmented MDP, yieldi…
A commonly expressed concern about the rise of the peer-to-peer rental market Airbnb is that hosts---those renting out their properties---impose costs on their unwitting neighbors. I consider the question of whether apartment building owners will, in a competitive rental market, set a building-specific Airbnb hosting p…
Paper presents a new policy gradient theorem using weak derivatives for reinforcement learning.
problem Continuous state-action reinforcement learning problems.
method Introduced an alternative policy gradient theorem using weak derivatives.
result The new approach yields algorithms that converge almost surely to stationary points of the value function.
In this paper, we propose a novel policy iteration method, called dynamic policy programming (DPP), to estimate the optimal policy in the infinite-horizon Markov decision processes. We prove the finite-iteration and asymptotic l\infty-norm performance-loss bounds for DPP in the presence of approximation/estimation erro…
In the field of reinforcement learning there has been recent progress towards safety and high-confidence bounds on policy performance. However, to our knowledge, no practical methods exist for determining high-confidence policy performance bounds in the inverse reinforcement learning setting---where the true reward fun…
To ensure stability of learning, state-of-the-art generalized policy iteration algorithms augment the policy improvement step with a trust region constraint bounding the information loss. The size of the trust region is commonly determined by the Kullback-Leibler (KL) divergence, which not only captures the notion of d…
Delusional bias is a fundamental source of error in approximate Q-learning. To date, the only techniques that explicitly address delusion require comprehensive search using tabular value estimates. In this paper, we develop efficient methods to mitigate delusional bias by training Q-approximators with labels that are "…
A risk-aware RL approach using RDEU and Wasserstein ball for robust performance.
problem Optimizing risk-aware performance criteria in uncertain environments.
method Rank dependent expected utility (RDEU) for risk assessment, Wasserstein ball for robustness, actor/agent framework.
result Explicit policy gradient formulae for robust optimization.
Transfer reinforcement learning (RL) aims at improving the learning efficiency of an agent by exploiting knowledge from other source agents trained on relevant tasks. However, it remains challenging to transfer knowledge between different environmental dynamics without having access to the source environments. In this …
We propose a probabilistic framework to directly insert prior knowledge in reinforcement learning (RL) algorithms by defining the behaviour policy as a Bayesian posterior distribution. Such a posterior combines task specific information with prior knowledge, thus allowing to achieve transfer learning across tasks. The …
Temporal point process is an expressive tool for modeling event sequences over time. In this paper, we take a reinforcement learning view whereby the observed sequences are assumed to be generated from a mixture of latent policies. The purpose is to cluster the sequences with different temporal patterns into the underl…
Unified policy controls diverse agents through modular neural networks.
problem Learning control policies for various agent morphologies.
method Shared Modular Policies (SMP) with decentralized control and message passing.
result A single modular policy controls multiple agent morphologies.
We propose a method for efficient training of Q-functions for continuous-state Markov Decision Processes (MDPs) such that the traces of the resulting policies satisfy a given Linear Temporal Logic (LTL) property. LTL, a modal logic, can express a wide range of time-dependent logical properties (including "safety") that…
Model-free deep reinforcement learning has been shown to exhibit good performance in domains ranging from video games to simulated robotic manipulation and locomotion. However, model-free methods are known to perform poorly when the interaction time with the environment is limited, as is the case for most real-world ro…
New algorithm reduces bias in off-policy reinforcement learning.
problem Challenges in designing off-policy reinforcement learning algorithms.
method Doubly robust off-policy actor-critic (DR-Off-PAC) with a single timescale structure.
result Establishes the first overall sample complexity analysis for a single time-scale off-policy AC algorithm.