ICLR 2023 论文列表

Although deep reinforcement learning (DRL) has many success stories, the large-scale deployment of policies learned through these advanced techniques in safety-critical scenarios is hindered by their lack of formal guarantees. Variational Markov Decision Processes (VAE-MDPs) are discrete latent space models that provide a reliable framework for distilling formally verifiable controllers from any RL policy. While the related guarantees address relevant practical aspects such as the satisfaction of performance and safety properties, the VAE approach suffers from several learning flaws (posterior collapse, slow learning speed, poor dynamics estimates), primarily due to the absence of abstraction and representation guarantees to support latent optimization. We introduce the Wasserstein auto-encoded MDP (WAE-MDP), a latent space model that fixes those issues by minimizing a penalized form of the optimal transport between the behaviors of the agent executing the original policy and the distilled policy, for which the formal guarantees apply. Our approach yields bisimulation guarantees while learning the distilled policy, allowing concrete optimization of the abstraction and representation model quality. Our experiments show that, besides distilling policies up to 10 times faster, the latent model quality is indeed better in general. Moreover, we present experiments from a simple time-to-failure verification algorithm on the latent space. The fact that our approach enables such simple verification techniques highlights its applicability.

Despite the recent advancement in multi-agent reinforcement learning (MARL), the MARL agents easily overfit the training environment and perform poorly in evaluation scenarios where other agents behave differently. Obtaining generalizable policies for MARL agents is thus necessary but challenging mainly due to complex multi-agent interactions. In this work, we model the MARL problem with Markov Games and propose a simple yet effective method, called ranked policy memory (RPM), i.e., to maintain a look-up memory of policies to achieve good generalizability. The main idea of RPM is to train MARL policies via gathering massive multi-agent interaction data. In particular, we first rank each agent’s policies by its training episode return, i.e., the episode return of each agent in the training environment; we then save the ranked policies in the memory; when an episode starts, each agent can randomly select a policy from the RPM as the behavior policy. Each agent uses the behavior policy to gather multi-agent interaction data for MARL training. This innovative self-play framework guarantees the diversity of multi-agent interaction in the training data. Experimental results on Melting Pot demonstrate that RPM enables MARL agents to interact with unseen agents in multi-agent generalization evaluation scenarios and complete given tasks. It significantly boosts the performance up to 818% on average.

Long-term engagement is preferred over immediate engagement in sequential recommendation as it directly affects product operational metrics such as daily active users (DAUs) and dwell time. Meanwhile, reinforcement learning (RL) is widely regarded as a promising framework for optimizing long-term engagement in sequential recommendation. However, due to expensive online interactions, it is very difficult for RL algorithms to perform state-action value estimation, exploration and feature extraction when optimizing long-term engagement. In this paper, we propose ResAct which seeks a policy that is close to, but better than, the online-serving policy. In this way, we can collect sufficient data near the learned policy so that state-action values can be properly estimated, and there is no need to perform online exploration. ResAct optimizes the policy by first reconstructing the online behaviors and then improving it via a Residual Actor. To extract long-term information, ResAct utilizes two information-theoretical regularizers to confirm the expressiveness and conciseness of features. We conduct experiments on a benchmark dataset and a large-scale industrial dataset which consists of tens of millions of recommendation requests. Experimental results show that our method significantly outperforms the state-of-the-art baselines in various long-term engagement optimization tasks.

The rise of generalist large-scale models in natural language and vision has made us expect that a massive data-driven approach could achieve broader generalization in other domains such as continuous control. In this work, we explore a method for learning a single policy that manipulates various forms of agents to solve various tasks by distilling a large amount of proficient behavioral data. In order to align input-output (IO) interface among multiple tasks and diverse agent morphologies while preserving essential 3D geometric relations, we introduce morphology-task graph, which treats observations, actions and goals/task in a unified graph representation. We also develop MxT-Bench for fast large-scale behavior generation, which supports procedural generation of diverse morphology-task combinations with a minimal blueprint and hardware-accelerated simulator. Through efficient representation and architecture selection on MxT-Bench, we find out that a morphology-task graph representation coupled with Transformer architecture improves the multi-task performances compared to other baselines including recent discrete tokenization, and provides better prior knowledge for zero-shot transfer or sample efficiency in downstream multi-task imitation learning. Our work suggests large diverse offline datasets, unified IO representation, and policy representation and architecture selection through supervised learning form a promising approach for studying and advancing morphology-task generalization.

Model-based reinforcement learning is one approach to increase sample efficiency. However, the accuracy of the dynamics model and the resulting compounding error over modelled trajectories are commonly regarded as key limitations. A natural question to ask is: How much more sample efficiency can be gained by improving the learned dynamics models? Our paper empirically answers this question for the class of model-based value expansion methods in continuous control problems. Value expansion methods should benefit from increased model accuracy by enabling longer rollout horizons and better value function approximations. Our empirical study, which leverages oracle dynamics models to avoid compounding model errors, shows that (1) longer horizons increase sample efficiency, but the gain in improvement decreases with each additional expansion step, and (2) the increased model accuracy only marginally increases the sample efficiency compared to learned models with identical horizons. Therefore, longer horizons and increased model accuracy yield diminishing returns in terms of sample efficiency. These improvements in sample efficiency are particularly disappointing when compared to model-free value expansion methods. Even though they introduce no computational overhead, we find their performance to be on-par with model-based value expansion methods. Therefore, we conclude that the limitation of model-based value expansion methods is not the model accuracy of the learned models. While higher model accuracy is beneficial, our experiments show that even a perfect model will not provide an un-rivaled sample efficiency but that the bottleneck lies elsewhere.

The spontaneous exchange of turns is a central aspect of human communication. Although turn-taking conventions come to us naturally, artificial dialogue agents struggle to coordinate, and must rely on hard-coded rules to engage in interactive conversations with human interlocutors. In this paper, we investigate the conditions under which artificial agents may naturally develop turn-taking conventions in a simple language game. We describe a cooperative task where success is contingent on the exchange of information along a shared communication channel where talking over each other hinders communication. Despite these environmental constraints, neural-network based agents trained to solve this task with reinforcement learning do not systematically adopt turn-taking conventions. However, we find that agents that do agree on turn-taking protocols end up performing better. Moreover, agents that are forced to perform turn-taking can learn to solve the task more quickly. This suggests that turn-taking may help to generate conversations that are easier for speakers to interpret.

Active learning has demonstrated data efficiency in many fields. Existing active learning algorithms, especially in the context of batch-mode deep Bayesian active models, rely heavily on the quality of uncertainty estimations of the model, and are often challenging to scale to large batches. In this paper, we propose Batch-BALanCe, a scalable batch-mode active learning algorithm, which combines insights from decision-theoretic active learning, combinatorial information measure, and diversity sampling. At its core, Batch-BALanCe relies on a novel decision-theoretic acquisition function that facilitates differentiation among different equivalence classes. Intuitively, each equivalence class consists of hypotheses (e.g., posterior samples of deep neural networks) with similar predictions, and Batch-BALanCe adaptively adjusts the size of the equivalence classes as learning progresses. To scale up the computation of queries to large batches, we further propose an efficient batch-mode acquisition procedure, which aims to maximize a novel combinatorial information measure defined through the acquisition function. We show that our algorithm can effectively handle realistic multi-class classification tasks, and achieves compelling performance on several benchmark datasets for active learning under both low- and large-batch regimes.

The exploration problem is one of the main challenges in deep reinforcement learning (RL). Recent promising works tried to handle the problem with population-based methods, which collect samples with diverse behaviors derived from a population of different exploratory policies. Adaptive policy selection has been adopted for behavior control. However, the behavior selection space is largely limited by the predefined policy population, which further limits behavior diversity. In this paper, we propose a general framework called Learnable Behavioral Control (LBC) to address the limitation, which a) enables a significantly enlarged behavior selection space via formulating a hybrid behavior mapping from all policies; b) constructs a unified learnable process for behavior selection. We introduce LBC into distributed off-policy actor-critic methods and achieve behavior control via optimizing the selection of the behavior mappings with bandit-based meta-controllers. Our agents have achieved 10077.52% mean human normalized score and surpassed 24 human world records within 1B training frames in the Arcade Learning Environment, which demonstrates our significant state-of-the-art (SOTA) performance without degrading the sample efficiency.

Deep Reinforcement Learning (Deep RL) and Evolutionary Algorithm (EA) are two major paradigms of policy optimization with distinct learning principles, i.e., gradient-based v.s. gradient-free. An appealing research direction is integrating Deep RL and EA to devise new methods by fusing their complementary advantages. However, existing works on combining Deep RL and EA have two common drawbacks:1) the RL agent and EA agents learn their policies individually, neglecting efficient sharing of useful common knowledge; 2) parameter-level policy optimization guarantees no semantic level of behavior evolution for the EA side. In this paper, we propose Evolutionary Reinforcement Learning with Two-scale State Representation and Policy Representation (ERL-Re$^2$), a novel solution to the aforementioned two drawbacks. The key idea of ERL-Re$^2$ is two-scale representation: all EA and RL policies share the same nonlinear state representation while maintaining individual linear policy representations. The state representation conveys expressive common features of the environment learned by all the agents collectively; the linear policy representation provides a favorable space for efficient policy optimization, where novel behavior-level crossover and mutation operations can be performed. Moreover, the linear policy representation allows convenient generalization of policy fitness with the help of Policy-extended Value Function Approximator (PeVFA), further improving the sample efficiency of fitness estimation. The experiments on a range of continuous control tasks show that ERL-Re$^2$ consistently outperforms advanced baselines and achieves the State Of The Art (SOTA). Our code is available on https://github.com/yeshenpy/ERL-Re2.

In this paper we study online Reinforcement Learning (RL) in partially observable dynamical systems. We focus on the Predictive State Representations (PSRs) model, which is an expressive model that captures other well-known models such as Partially Observable Markov Decision Processes (POMDP). PSR represents the states using a set of predictions of future observations and is defined entirely using observable quantities. We develop a novel model-based algorithm for PSRs that can learn a near optimal policy in sample complexity scaling polynomially with respect to all the relevant parameters of the systems. Our algorithm naturally works with function approximation to extend to systems with potentially large state and observation spaces. We show that given a realizable model class, the sample complexity of learning the near optimal policy only scales polynomially with respect to the statistical complexity of the model class, without any explicit polynomial dependence on the size of the state and observation spaces. Notably, our work is the first work that shows polynomial sample complexities to compete with the globally optimal policy in PSRs. Finally, we demonstrate how our general theorem can be directly used to derive sample complexity bounds for special models including $m$-step weakly revealing and $m$-step decodable tabular POMDPs, POMDPs with low-rank latent transition, and POMDPs with linear emission and latent transition.

Representation learning often plays a critical role in avoiding the curse of dimensionality in reinforcement learning. A representative class of algorithms exploits spectral decomposition of the stochastic transition dynamics to construct representations that enjoy strong theoretical properties in idealized settings. However, current spectral methods suffer from limited applicability because they are constructed for state-only aggregation and are derived from a policy-dependent transition kernel, without considering the issue of exploration. To address these issues, we propose an alternative spectral method, Spectral Decomposition Representation (SPEDER), that extracts a state-action abstraction from the dynamics without inducing spurious dependence on the data collection policy, while also balancing the exploration-versus-exploitation trade-off during learning. A theoretical analysis establishes the sample efficiency of the proposed algorithm in both the online and offline settings. In addition, an experimental investigation demonstrates superior performance over current state-of-the-art algorithms across several RL benchmarks.

No-press Diplomacy is a complex strategy game involving both cooperation and competition that has served as a benchmark for multi-agent AI research. While self-play reinforcement learning has resulted in numerous successes in purely adversarial games like chess, Go, and poker, self-play alone is insufficient for achieving optimal performance in domains involving cooperation with humans. We address this shortcoming by first introducing a planning algorithm we call DiL-piKL that regularizes a reward-maximizing policy toward a human imitation-learned policy. We prove that this is a no-regret learning algorithm under a modified utility function. We then show that DiL-piKL can be extended into a self-play reinforcement learning algorithm we call RL-DiL-piKL that provides a model of human play while simultaneously training an agent that responds well to this human model. We used RL-DiL-piKL to train an agent we name Diplodocus. In a 200-game no-press Diplomacy tournament involving 62 human participants spanning skill levels from beginner to expert, two Diplodocus agents both achieved a higher average score than all other participants who played more than two games, and ranked first and third according to an Elo ratings model.

Recently, the Successor Features and Generalized Policy Improvement (SF&GPI) framework has been proposed as a method for learning, composing and transferring predictive knowledge and behavior. SF&GPI works by having an agent learn predictive representations (SFs) that can be combined for transfer to new tasks with GPI. However, to be effective this approach requires state features that are useful to predict, and these state-features are typically hand-designed. In this work, we present a novel neural network architecture, “Modular Successor Feature Approximators” (MSFA), where modules both discover what is useful to predict, and learn their own predictive representations. We show that MSFA is able to better generalize compared to baseline architectures for learning SFs and a modular network that discovers factored state representations.

This work studies an algorithm, which we call magnetic mirror descent, that is inspired by mirror descent and the non-Euclidean proximal gradient algorithm. Our contribution is demonstrating the virtues of magnetic mirror descent as both an equilibrium solver and as an approach to reinforcement learning in two-player zero-sum games. These virtues include: 1) Being the first quantal response equilibria solver to achieve linear convergence for extensive-form games with first order feedback; 2) Being the first standard reinforcement learning algorithm to achieve empirically competitive results with CFR in tabular settings; 3) Achieving favorable performance in 3x3 Dark Hex and Phantom Tic-Tac-Toe as a self-play deep reinforcement learning algorithm.

Training deep reinforcement learning (DRL) models usually requires high computation costs. Therefore, compressing DRL models possesses immense potential for training acceleration and model deployment. However, existing methods that generate small models mainly adopt the knowledge distillation-based approach by iteratively training a dense network. As a result, the training process still demands massive computing resources. Indeed, sparse training from scratch in DRL has not been well explored and is particularly challenging due to non-stationarity in bootstrap training. In this work, we propose a novel sparse DRL training framework, “the Rigged Reinforcement Learning Lottery” (RLx2), which builds upon gradient-based topology evolution and is capable of training a sparse DRL model based entirely on a sparse network. Specifically, RLx2 introduces a novel multi-step TD target mechanism with a dynamic-capacity replay buffer to achieve robust value learning and efficient topology exploration in sparse models. It also reaches state-of-the-art sparse training performance in several tasks, showing $7.5\times$-$20\times$ model compression with less than $3\%$ performance degradation and up to $20\times$ and $50\times$ FLOPs reduction for training and inference, respectively.

Future- or return-conditioned supervised learning is an emerging paradigm for offline reinforcement learning (RL), in which the future outcome (i.e., return) associated with a sequence of actions in an offline dataset is used as input to a policy trained to imitate those same actions. While return-conditioning is at the heart of popular algorithms such as decision transformer (DT), these methods tend to perform poorly in highly stochastic environments, where an occasional high return associated with a sequence of actions may be due more to the randomness of the environment than to the actions themselves. Such situations can lead to a learned policy that is inconsistent with its conditioning inputs; i.e., using the policy – while conditioned on a specific desired return – to act in the environment can lead to a distribution of real returns that is wildly different than desired. In this work, we propose the dichotomy of control (DoC), a future-conditioned supervised learning framework that separates mechanisms within a policy’s control (actions) from those outside of a policy’s control (environment stochasticity). We achieve this by conditioning the policy on a latent variable representation of the future and designing a mutual information constraint that removes any future information from the latent variable that is only due to randomness of the environment. Theoretically, we show that DoC yields policies that are consistent with their conditioning inputs, ensuring that conditioning a learned policy on a desired high-return future outcome will correctly induce high-return behavior. Empirically, we show that DoC is able to achieve significantly better performance than DT on environments with highly stochastic rewards (e.g., Bandit) and transitions (e.g., FrozenLake).

Reinforcement learning (RL) in text-based games has developed rapidly and achieved promising results. However, little effort has been expended to design agents that pursue objectives while behaving morally, which is a critical issue in the field of autonomous agents. In this paper, we propose a general framework named Moral Awareness Adaptive Learning (MorAL) that enhances the morality capacity of an agent using a plugin moral-aware learning model. The framework allows the agent to execute task learning and morality learning adaptively. The agent selects trajectories from past experiences during task learning. Meanwhile, the trajectories are used to conduct self-imitation learning with a moral-enhanced objective. In order to achieve the trade-off between morality and task progress, the agent uses the combination of task policy and moral policy for action selection. We evaluate on the Jiminy Cricket benchmark, a set of text-based games with various scenes and dense morality annotations. Our experiments demonstrate that, compared with strong contemporary value alignment approaches, the proposed framework improves task performance while reducing immoral behaviours in various games.

Humans manipulate various kinds of fluids in their everyday life: creating latte art, scooping floating objects from water, rolling an ice cream cone, etc. Using robots to augment or replace human labors in these daily settings remain as a challenging task due to the multifaceted complexities of fluids. Previous research in robotic fluid manipulation mostly consider fluids governed by an ideal, Newtonian model in simple task settings (e.g., pouring water into a container). However, the vast majority of real-world fluid systems manifest their complexities in terms of the fluid’s complex material behaviors (e.g., elastoplastic deformation) and multi-component interactions (e.g. coffee and frothed milk when making latte art), both of which were well beyond the scope of the current literature. To evaluate robot learning algorithms on understanding and interacting with such complex fluid systems, a comprehensive virtual platform with versatile simulation capabilities and well-established tasks is needed. In this work, we introduce FluidLab, a simulation environment with a diverse set of manipulation tasks involving complex fluid dynamics. These tasks address interactions between solid and fluid as well as among multiple fluids. At the heart of our platform is a fully differentiable physics simulator, FluidEngine, providing GPU-accelerated simulations and gradient calculations for various material types and their couplings, extending the scope of the existing differentiable simulation engines. We identify several challenges for fluid manipulation learning by evaluating a set of reinforcement learning and trajectory optimization methods on our platform. To address these challenges, we propose several domain-specific optimization schemes coupled with differentiable physics, which are empirically shown to be effective in tackling optimization problems featured by fluid system’s non-convex and non-smooth properties. Furthermore, we demonstrate reasonable sim-to-real transfer by deploying optimized trajectories in real-world settings. FluidLab is publicly available at: https://fluidlab2023.github.io.

This paper presents a multi-agent reinforcement learning (MARL) scheme for proactive Multi-Camera Collaboration in 3D Human Pose Estimation in dynamic human crowds. Traditional fixed-viewpoint multi-camera solutions for human motion capture (MoCap) are limited in capture space and susceptible to dynamic occlusions. Active camera approaches proactively control camera poses to find optimal viewpoints for 3D reconstruction. However, current methods still face challenges with credit assignment and environment dynamics. To address these issues, our proposed method introduces a novel Collaborative Triangulation Contribution Reward (CTCR) that improves convergence and alleviates multi-agent credit assignment issues resulting from using 3D reconstruction accuracy as the shared reward. Additionally, we jointly train our model with multiple world dynamics learning tasks to better capture environment dynamics and encourage anticipatory behaviors for occlusion avoidance. We evaluate our proposed method in four photo-realistic UE4 environments to ensure validity and generalizability. Empirical results show that our method outperforms fixed and active baselines in various scenarios with different numbers of cameras and humans.