Study agnostic RL in large state spaces with weak function approximation.
problem Statistical intractability of agnostic policy learning in various environments.
method Investigates agnostic policy learning with different forms of environment access.
result Agnostic policy learning remains statistically intractable with certain forms of environment access.
This paper proposes a system-agnostic policy for dynamic scheduling.
problem Dynamic scheduling in changing systems is challenging due to system-specific optimal policies.
method Descriptive policy that learns a system-agnostic scheduling principle.
result System-agnostic meta-learning enables adaptation to unseen system characteristics.
New insights into RL with weak function approximation.
problem Statistical complexity of RL with function approximation in large state spaces.
method Agnostic policy learning framework, exploring environment access, coverage, and representational conditions.
result Characterization of fundamental performance bounds and statistical separations.
In this paper we target the problem of transferring policies across multiple environments with different dynamics parameters and motor noise variations, by introducing a framework that decouples the processes of policy learning and system identification. Efficiently transferring learned policies to an unknown environme…
New complexity measure helps in agnostic reinforcement learning with or without access to MDP dynamics.
problem Understanding the number of rounds needed to learn an ε-suboptimal policy in unknown MDPs.
method Introducing spanning capacity as a new complexity measure and developing POPLER algorithm.
result There is a separation between generative and online access models for agnostic learnability.
New RL algorithms learn policies competitive with best in class without assuming optimal policy.
problem Agnostic RL where best policy not necessarily optimal.
method First-order optimization in non-Euclidean space, reducing to policy learning.
result Sample complexity upper bounds for three algorithms under VGD condition.
New method improves convergence of RL meta-learning.
problem Improving convergence in model-agnostic meta-reinforcement learning.
method Proposes Stochastic Gradient Meta-Reinforcement Learning (SG-MRL) to find ε-first-order stationary points. result Derives iteration and sample complexity for SG-MRL.
New algorithm finds optimal policies without knowing reward functions.
problem Reward-agnostic exploration in reinforcement learning.
method Designs an algorithm that explores without reward information, achieving minimax optimality.
result Achieves provable minimax optimality in finding optimal policies for multiple reward functions.
Paper proposes a policy-search algorithm to learn entropy-maximizing exploration policies in reward-free environments.
problem Reward-free learning in high-dimensional, continuous-control domains.
method Maximum Entropy POLicy optimization (MEPOL) algorithm that maximizes a non-parametric state entropy estimate.
result MEPOL learns a maximum-entropy exploration policy that facilitates learning various reward-based tasks.
The composition of elementary behaviors to solve challenging transfer learning problems is one of the key elements in building intelligent machines. To date, there has been plenty of work on learning task-specific policies or skills but almost no focus on composing necessary, task-agnostic skills to find a solution to …
Study agnostic feature-based dynamic pricing models with linear policies and noisy valuations.
problem Tackles dynamic pricing with unknown noise and no assumptions on data.
method Studies two agnostic models: linear policy and linear noisy valuation, presenting algorithms and regret bounds.
result Demonstrates no-regret learning is possible under weak assumptions, but noisy feedback is not significantly more useful than bandit feedback.
New method learns robot actions from videos without explicit labels.
problem Training robots to perform tasks from few demonstrations.
method Uses images and text for task-agnostic and general representation, synthesizes hallucinated actions, and applies dense correspondences.
result Trains robot policies solely from RGB videos, achieving diverse tasks across different robots and environments.
Generically learns movement control policies from exploration data.
problem Movement optimization in physically based characters.
method Parameterizes actions as target states, learns low-level control policy.
result Improves movement optimization across multiple tasks and algorithms.
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.
While model-based deep reinforcement learning (RL) holds great promise for sample efficiency and generalization, learning an accurate dynamics model is often challenging and requires substantial interaction with the environment. A wide variety of domains have dynamics that share common foundations like the laws of clas…
In many real-world reinforcement learning applications, access to the environment is limited to a fixed dataset, instead of direct (online) interaction with the environment. When using this data for either evaluation or training of a new policy, accurate estimates of discounted stationary distribution ratios -- correct…
Proposes PIC and POIC for measuring task difficulty in RL.
problem Lack of metrics to measure task difficulty in RL.
method Introduces policy information capacity (PIC) and policy-optimal information capacity (POIC) as metrics based on mutual information.
result Empirically shows PIC and POIC correlate with task solvability better than alternatives.
Paper proposes a hybrid RL algorithm that combines offline and online data without needing reward info.
problem How to efficiently use online data to improve RL policies using only offline data.
method A three-stage hybrid RL algorithm that uses reward-agnostic exploration and model-based offline RL.
result The hybrid RL algorithm outperforms both pure offline and pure online RL in sample complexity.
While recent progress has spawned very powerful machine learning systems, those agents remain extremely specialized and fail to transfer the knowledge they gain to similar yet unseen tasks. In this paper, we study a simple reinforcement learning problem and focus on learning policies that encode the proper invariances …
Inverse reinforcement learning (IRL) enables an agent to learn complex behavior by observing demonstrations from a (near-)optimal policy. The typical assumption is that the learner's goal is to match the teacher's demonstrated behavior. In this paper, we consider the setting where the learner has its own preferences th…
New algorithm finds optimal policy with polynomial trajectories in deterministic systems.
problem Finding optimal policy in deterministic systems with function approximation.
method Novel recursion-based algorithm with tight bounds on error and sample complexity.
result Optimal policy found using O(dimE) trajectories with $δ= O\left(ρ/\sqrt{\dim_E}
ight)$. Task-agnostic RL tackles exploration in MDPs with multiple tasks.
problem Challenges in reinforcement learning with multiple tasks or conflicting objectives.
method Task-agnostic RL framework, UCBZero algorithm.
result UCBZero finds near-optimal policies for multiple tasks efficiently.
We propose a method for tackling catastrophic forgetting in deep reinforcement learning that is \textit{agnostic} to the timescale of changes in the distribution of experiences, does not require knowledge of task boundaries, and can adapt in \textit{continuously} changing environments. In our \textit{policy consolidati…
We review basic concepts of convex duality, focusing on the very general and supremely useful Fenchel-Rockafellar duality. We summarize how this duality may be applied to a variety of reinforcement learning (RL) settings, including policy evaluation or optimization, online or offline learning, and discounted or undisco…
A novel framework uses goal-conditioned reinforcement learning to generate diverse samples.
problem Generating high-quality, diverse samples from generative models.
method Two agents: GC-agent learns to reconstruct the training set, S-agent learns to imitate GC-agent without knowing the goals.
result Empirically, the method generates diverse and high-quality samples in image synthesis.
New approach makes deep reinforcement learning robust without assuming adversary knowledge.
problem Deep reinforcement learning policies are vulnerable to state observation perturbations.
method Proposes an adversary agnostic robust DRL paradigm using policy distillation with two terms: prescription gap maximization and Jacobian regularization.
result Boosts adversarial robustness on five Atari games compared to state-of-the-art methods.
Traditional reinforcement learning agents learn from experience, past or present, gained through interaction with their environment. Our approach synthesizes experience, without requiring an agent to interact with their environment, by asking the policy directly "Are there situations X, Y, and Z, such that in these sit…
CoinDICE estimates confidence intervals for unknown behavior policies in reinforcement learning.
problem Estimating value of a target policy using only behavior policy data.
method Function space embedding, generalized empirical likelihood method, Lagrangian optimization.
result Valid confidence intervals with tighter and more accurate estimates than existing methods.
A fundamental issue in reinforcement learning algorithms is the balance between exploration of the environment and exploitation of information already obtained by the agent. Especially, exploration has played a critical role for both efficiency and efficacy of the learning process. However, Existing works for explorati…
Unified DICE estimators as regularized Lagrangians for improved off-policy evaluation.
problem Improving off-policy evaluation from behavior-agnostic data.
method Unified derivation of DICE estimators as regularized Lagrangians of a linear program.
result Dual solutions offer greater flexibility and provide superior estimates in practice.
This study optimizes offline reinforcement learning methods for various tasks without rewards.
problem Optimizing offline reinforcement learning for multiple tasks without rewards.
method Designing a new model-based approach with singleton absorbing MDPs to achieve optimal convergence rates.
result Achieved optimal convergence rates for offline reinforcement learning in various settings.
Proposes a new theoretical framework for PbRL that requires less human feedback.
problem Lack of theoretical work capturing practical PbRL frameworks.
method Introduces a reward-agnostic PbRL framework that acquires exploratory trajectories before human feedback.
result Demonstrates improved sample complexity for learning optimal policies in linear and low-rank MDPs.
Algorithm learns to switch control among agents in a team.
problem Learning to switch control among reinforcement learning agents.
method 2-layer Markov decision process, upper confidence bounds, shared confidence bounds.
result Sublinear total regret with shared confidence bounds.
VRER selectively reuses past observations to reduce variance in policy optimization.
problem Lack of effective experience replay for accelerating policy optimization in complex systems.
method Variance Reduction Experience Replay (VRER) framework that selectively reuses informative samples.
result VRER reduces gradient variance and improves policy learning over state-of-the-art algorithms.
Model-agnostic meta-learning (MAML) is a meta-learning technique to train a model on a multitude of learning tasks in a way that primes the model for few-shot learning of new tasks. The MAML algorithm performs well on few-shot learning problems in classification, regression, and fine-tuning of policy gradients in reinf…
We study the off-policy evaluation problem---estimating the value of a target policy using data collected by another policy---under the contextual bandit model. We consider the general (agnostic) setting without access to a consistent model of rewards and establish a minimax lower bound on the mean squared error (MSE).…
Energy-efficient DL inference for IoT devices reduces power consumption and improves performance.
problem Energy inefficiency in deep learning models for IoT devices.
method Energy-aware early exiting policy to balance energy consumption and inference accuracy.
result Accuracy and service rate improved up to 25% and 35% respectively.
New algorithm reduces best-in-class regret in contextual bandits.
problem Compete with the best policy in a class without model restrictions.
method Proposes an algorithm that updates policies by minimizing a pessimistic objective, including a clipped inverse-propensity estimate and variance penalty.
result Achieves fast best-in-class regret rates, including polylogarithmic rates in the parametric case.
A fundamental problem in control is to learn a model of a system from observations that is useful for controller synthesis. To provide good performance guarantees, existing methods must assume that the real system is in the class of models considered during learning. We present an iterative method with strong guarantee…
Meta-reinforcement learning improves fault-adaptive control efficiency.
problem Adaptive control under abrupt system faults with strict time constraints.
method Model-agnostic meta learning (MAML) with a fault library of prior policies.
result Improved sample efficiency and quick adaptation to new faults.
Meta-Reinforcement learning approaches aim to develop learning procedures that can adapt quickly to a distribution of tasks with the help of a few examples. Developing efficient exploration strategies capable of finding the most useful samples becomes critical in such settings. Existing approaches towards finding effic…
CODA resolves coordination issues in offline multi-agent reinforcement learning.
problem Coordination failure in offline multi-agent reinforcement learning.
method Diffusion-based multi-agent trajectory generator for data augmentation.
result CODA resolves coordination pathologies in continuous polynomial games and complex benchmarks.
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 …
New method optimizes policies without assuming known link functions between preferences and rewards.
problem Policy alignment with unknown and unrestricted link functions.
method Formulates an f-divergence-constrained reward maximization problem, learning policies directly. result Induces a semiparametric single-index binary choice model for policy alignment.
RL learns to ignore factors in factor investing portfolios.
problem Combining factor investing and reinforcement learning for optimal portfolio allocation.
method RL agent learns through sequential allocations based on firms' characteristics using Dirichlet distributions.
result RL-based portfolios are very close to equally-weighted allocations, indicating agnostic factor learning.
New framework learns policies for partially observable systems.
problem Learning policies in partially observable dynamical systems.
method Partially Observable Bilinear Actor-Critic framework.
result Algorithm can learn against optimal policies in certain cases.
We consider off-policy policy evaluation when the trajectory data are generated by multiple behavior policies. Recent work has shown the key role played by the state or state-action stationary distribution corrections in the infinite horizon context for off-policy policy evaluation. We propose estimated mixture policy …
The growing prospect of deep reinforcement learning (DRL) being used in cyber-physical systems has raised concerns around safety and robustness of autonomous agents. Recent work on generating adversarial attacks have shown that it is computationally feasible for a bad actor to fool a DRL policy into behaving sub optima…