A prominent challenge of offline reinforcement learning (RL) is the issue of hidden confounding: unobserved variables may influence both the actions taken by the agent and the observed outcomes. Hidden confounding can compromise the validity of any causal conclusion drawn from data and presents a major obstacle to effective offline RL. In the present paper, we tackle the problem of hidden confounding in the nonidentifiable setting. We propose a definition of uncertainty due to hidden confounding bias, termed delphic uncertainty, which uses variation over world models compatible with the observations, and differentiate it from the well-known epistemic and aleatoric uncertainties. We derive a practical method for estimating the three types of uncertainties, and construct a pessimistic offline RL algorithm to account for them. Our method does not assume identifiability of the unobserved confounders, and attempts to reduce the amount of confounding bias. We demonstrate through extensive experiments and ablations the efficacy of our approach on a sepsis management benchmark, as well as on electronic health records. Our results suggest that nonidentifiable hidden confounding bias can be mitigated to improve offline RL solutions in practice.

Recent advances in visual reinforcement learning (RL) have led to impressive success in handling complex tasks. However, these methods have demonstrated limited generalization capability to visual disturbances, which poses a significant challenge for their real-world application and adaptability. Though normalization techniques have demonstrated huge success in supervised and unsupervised learning, their applications in visual RL are still scarce. In this paper, we explore the potential benefits of integrating normalization into visual RL methods with respect to generalization performance. We find that, perhaps surprisingly, incorporating suitable normalization techniques is sufficient to enhance the generalization capabilities, without any additional special design. We utilize the combination of two normalization techniques, CrossNorm and SelfNorm, for generalizable visual RL. Extensive experiments are conducted on DMControl Generalization Benchmark and CARLA to validate the effectiveness of our method. We show that our method significantly improves generalization capability while only marginally affecting sample efficiency. In particular, when integrated with DrQ-v2, our method enhances the test performance of DrQ-v2 on CARLA across various scenarios, from 14% of the training performance to 97%.

Dynamical systems with complex behaviours, e.g. immune system cells interacting with a pathogen, are commonly modelled by splitting the behaviour into different regimes, or modes, each with simpler dynamics, and then learning the switching behaviour from one mode to another. Switching Dynamical Systems (SDS) are a powerful tool that automatically discovers these modes and mode-switching behaviour from time series data. While effective, these methods focus on independent objects, where the modes of one object are independent of the modes of the other objects. In this paper, we focus on the more general interacting object setting for switching dynamical systems, where the per-object dynamics also depends on an unknown and dynamically changing subset of other objects and their modes. To this end, we propose a novel graph-based approach for switching dynamical systems, GRAph Switching dynamical Systems (GRASS), in which we use a dynamic graph to characterize interactions between objects and learn both intra-object and inter-object mode-switching behaviour. We introduce two new datasets for this setting, a synthesized ODE-driven particles dataset and a real-world Salsa Couple Dancing dataset. Experiments show that GRASS can consistently outperforms previous state-of-the-art methods.

Recent offline meta-reinforcement learning (meta-RL) methods typically utilize task-dependent behavior policies (e.g., training RL agents on each individual task) to collect a multi-task dataset. However, these methods always require extra information for fast adaptation, such as offline context for testing tasks. To address this problem, we first formally characterize a unique challenge in offline meta-RL: transition-reward distribution shift between offline datasets and online adaptation. Our theory finds that out-of-distribution adaptation episodes may lead to unreliable policy evaluation and that online adaptation with in-distribution episodes can ensure adaptation performance guarantee. Based on these theoretical insights, we propose a novel adaptation framework, called In-Distribution online Adaptation with uncertainty Quantification (IDAQ), which generates in-distribution context using a given uncertainty quantification and performs effective task belief inference to address new tasks. We find a return-based uncertainty quantification for IDAQ that performs effectively. Experiments show that IDAQ achieves state-of-the-art performance on the Meta-World ML1 benchmark compared to baselines with/without offline adaptation.

Risk assessment is a crucial component of collision warning and avoidance systems in intelligent vehicles. To accurately detect potential vehicle collisions, reachability-based formal approaches have been developed to ensure driving safety, but suffer from over-conservatism, potentially leading to false-positive risk events in complicated real-world applications. In this work, we combine two reachability analysis techniques, i.e., backward reachable set (BRS) and stochastic forward reachable set (FRS), and propose an integrated probabilistic collision detection framework in highway driving. Within the framework, we can firstly use a BRS to formally check whether a two-vehicle interaction is safe; otherwise, a prediction-based stochastic FRS is employed to estimate a collision probability at each future time step. In doing so, the framework can not only identify non-risky events with guaranteed safety, but also provide accurate collision risk estimation in safety-critical events. To construct the stochastic FRS, we develop a neural network-based acceleration model for surrounding vehicles, and further incorporate confidence-aware dynamic belief to improve the prediction accuracy. Extensive experiments are conducted to validate the performance of the acceleration prediction model based on naturalistic highway driving data, and the efficiency and effectiveness of the framework with the infused confidence belief are tested both in naturalistic and simulated highway scenarios. The proposed risk assessment framework is promising in real-world applications.

Stackelberg equilibria arise naturally in a range of popular learning problems, such as in security games or indirect mechanism design, and have received increasing attention in the reinforcement learning literature. We present a general framework for implementing Stackelberg equilibria search as a multi-agent RL problem, allowing a wide range of algorithmic design choices. We discuss how previous approaches can be seen as specific instantiations of this framework. As a key insight, we note that the design space allows for approaches not previously seen in the literature, for instance by leveraging multitask and meta-RL techniques for follower convergence. We propose one such approach using contextual policies, and evaluate it experimentally on both standard and novel benchmark domains, showing greatly improved sample efficiency compared to previous approaches. Finally, we explore the effect of adopting algorithm designs outside the borders of our framework.

We examine online safe multi-agent reinforcement learning using constrained Markov games in which agents compete by maximizing their expected total rewards under a constraint on expected total utilities. Our focus is confined to an episodic two-player zero-sum constrained Markov game with independent transition functions that are unknown to agents, adversarial reward functions, and stochastic utility functions. For such a Markov game, we employ an approach based on the occupancy measure to formulate it as an online constrained saddle-point problem with an explicit constraint. We extend the Lagrange multiplier method in constrained optimization to handle the constraint by creating a generalized Lagrangian with minimax decision primal variables and a dual variable. Next, we develop an upper confidence reinforcement learning algorithm to solve this Lagrangian problem while balancing exploration and exploitation. Our algorithm updates the minimax decision primal variables via online mirror descent and the dual variable via projected gradient step and we prove that it enjoys sublinear rate $ O((|X|+|Y|) L \sqrt{T(|A|+|B|)}))$ for both regret and constraint violation after playing $T$ episodes of the game. Here, $L$ is the horizon of each episode, $(|X|,|A|)$ and $(|Y|,|B|)$ are the state/action space sizes of the min-player and the max-player, respectively. To the best of our knowledge, we provide the first provably efficient online safe reinforcement learning algorithm in constrained Markov games.

Centralized Training for Decentralized Execution, where agents are trained offline in a centralized fashion and execute online in a decentralized manner, has become a popular approach in Multi-Agent Reinforcement Learning (MARL). In particular, it has become popular to develop actor-critic methods that train decentralized actors with a centralized critic where the centralized critic is allowed access global information of the entire system, including the true system state. Such centralized critics are possible given offline information and are not used for online execution. While these methods perform well in a number of domains and have become a de facto standard in MARL, using a centralized critic in this context has yet to be sufficiently analyzed theoretically or empirically. In this paper, we therefore formally analyze centralized and decentralized critic approaches, and analyze the effect of using state-based critics in partially observable environments. We derive theories contrary to the common intuition: critic centralization is not strictly beneficial, and using state values can be harmful. We further prove that, in particular, state-based critics can introduce unexpected bias and variance compared to history-based critics. Finally, we demonstrate how the theory applies in practice by comparing different forms of critics on a wide range of common multi-agent benchmarks. The experiments show practical issues such as the difficulty of representation learning with partial observability, which highlights why the theoretical problems are often overlooked in the literature.

In this paper, we quantify the impact of using nonconvergent Markov chains to train Energy-Based EBMs offer several fundamental advantages over their competitors models (EBMs). In particular, we show analytically due to their simplicity: A single neural network is that EBMs trained with non-persistent short involved in training, which means that fewer parameters runs to estimate the gradient can perfectly reproduce need to be learned and training is less costly. They are also a set of empirical statistics of the data, not at particularly appealing for interpretive applications: Once the level of the equilibrium measure, but through trained, the energy function can be analyzed with statistical a precise dynamical process. Our results provide a mechanics tools (Decelle & Furtlehner, 2021b), or shallow first-principles explanation for the observations of EBMs can serve as an effective model to "learn" something recent works proposing the strategy of using short from the data. EBMs have been exploited for instance to runs starting from random initial conditions as an infer the three dimensional structure (Morcos et al., 2011) efficient way to generate high-quality samples in or building blocks (Tubiana et al., 2019) of proteins, to generate EBMs, and lay the groundwork for using EBMs artificial pieces of genome (Yelmen et al., 2021), for as diffusion models. After explaining this effect in neuroimaging (Hjelm et al., 2014), simulation of complex generic EBMs, we analyze two solvable models in wavefunctions in quantum many-body physics (Carleo & which the effect of the non-convergent sampling Troyer, 2017; Melko et al., 2019), or to impute missing in the trained parameters can be described in detail.

We propose a policy gradient algorithm for robust infinite-horizon Markov Decision Processes (MDPs) with non-rectangular uncertainty sets, thereby addressing an open challenge in the robust MDP literature. Indeed, uncertainty sets that display statistical optimality properties and make optimal use of limited data often fail to be rectangular. Unfortunately, the corresponding robust MDPs cannot be solved with dynamic programming techniques and are in fact provably intractable. This prompts us to develop a projected Langevin dynamics algorithm tailored to the robust policy evaluation problem, which offers global optimality guarantees. We also propose a deterministic policy gradient method that solves the robust policy evaluation problem approximately, and we prove that the approximation error scales with a new measure of non-rectangularity of the uncertainty set. Numerical experiments showcase that our projected Langevin dynamics algorithm can escape local optima, while algorithms tailored to rectangular uncertainty fail to do so.