In reinforcement learning, unsupervised skill discovery aims to learn diverse skills without extrinsic rewards. Previous methods discover skills by maximizing the mutual information (MI) between states and skills. However, such an MI objective tends to learn simple and static skills and may hinder exploration. In this paper, we propose a novel unsupervised skill discovery method through contrastive learning among behaviors, which makes the agent produce similar behaviors for the same skill and diverse behaviors for different skills. Under mild assumptions, our objective maximizes the MI between different behaviors based on the same skill, which serves as an upper bound of the previous MI objective. Meanwhile, our method implicitly increases the state entropy to obtain better state coverage. We evaluate our method on challenging mazes and continuous control tasks. The results show that our method generates diverse and far-reaching skills, and also obtains competitive performance in downstream tasks compared to the state-of-the-art methods.

Deploying reinforcement learning agents in the real world can be challenging due to the risks associated with learning through trial and error. We propose a task-agnostic method that leverages small sets of safe and unsafe demonstrations to improve the safety of RL agents during learning. The method compares the current trajectory of the agent with both sets of demonstrations at every step, and filters the trajectory if it resembles the unsafe demonstrations. We perform ablation studies on different filtering strategies and investigate the impact of the number of demonstrations on performance. Our method is compatible with any stand-alone RL algorithm and can be applied to any task. We evaluate our method on three tasks from OpenAI Gym's Mujoco benchmark and two state-of-the-art RL algorithms. The results demonstrate that our method significantly reduces the crash rate of the agent while converging to, and in most cases even improving, the performance of the stand-alone agent.

We apply reinforcement learning techniques to topic modeling by replacing the variational autoencoder in ProdLDA with a continuous action space reinforcement learning policy. We train the system with a policy gradient algorithm REINFORCE. Additionally, we introduced several modifications: modernize the neural network architecture, weight the ELBO loss, use contextual embeddings, and monitor the learning process via computing topic diversity and coherence for each training step. Experiments are performed on 11 data sets. Our unsupervised model outperforms all other unsupervised models and performs on par with or better than most models using supervised labeling. Our model is outperformed on certain data sets by a model using supervised labeling and contrastive learning. We have also conducted an ablation study to provide empirical evidence of performance improvements from changes we made to ProdLDA and found that the reinforcement learning formulation boosts performance.

Cooperative Multi-agent Reinforcement Learning (MARL) has attracted prominent attention in recent years [1], and achieved great progress in multiple aspects, like path finding [2], active voltage control [3], and dynamic algorithm configuration [4]. Among the multitudinous methods, researchers, on the one hand, focus on facilitating coordination ability via solving specific challenges, including non-stationarity [5], credit assignment [6], and scalability [7]. Other works, on the other hand, investigate the cooperative MARL from multiple aspects, like efficient communication [8], zero-shot coordination (ZSC) [9], policy robustness [10], etc. A lot of methods emerge as promising solutions for different scenarios, including policy-based ones [11,12], value-based series [13,14], and many other variants, showing remarkable coordination ability in a wide range of tasks like SMAC [15]. Despite the great success, the mainstream cooperative MARL methods are still restricted to being trained in one single task or multiple tasks simultaneously, assuming that the agents have access to data from all tasks at all times, which is unrealistic for physical agents in the real world that can only attend to one task at a time. Continual Reinforcement Learning plays a promising role in the mentioned problem [16], where the agent aims to avoid catastrophic forgetting, as well as enable knowledge transfer to new tasks (a.k.a.

In Reinforcement Learning (RL), an agent acts in an unknown environment to maximize the expected cumulative discounted sum of an external reward signal, i.e., the expected return. In practice, in many tasks of interest, such as policy optimization, the agent usually spends its interaction budget by collecting episodes of fixed length within a simulator (i.e., Monte Carlo simulation). However, given the discounted nature of the RL objective, this data collection strategy might not be the best option. Indeed, the rewards taken in early simulation steps weigh exponentially more than future rewards. Taking a cue from this intuition, in this paper, we design an a-priori budget allocation strategy that leads to the collection of trajectories of different lengths, i.e., truncated. The proposed approach provably minimizes the width of the confidence intervals around the empirical estimates of the expected return of a policy. After discussing the theoretical properties of our method, we make use of our trajectory truncation mechanism to extend Policy Optimization via Importance Sampling (POIS, Metelli et al., 2018) algorithm. Finally, we conduct a numerical comparison between our algorithm and POIS: the results are consistent with our theory and show that an appropriate truncation of the trajectories can succeed in improving performance.

When autonomous vehicles are deployed on public roads, they will encounter countless and diverse driving situations. Many manually designed driving policies are difficult to scale to the real world. Fortunately, reinforcement learning has shown great success in many tasks by automatic trial and error. However, when it comes to autonomous driving in interactive dense traffic, RL agents either fail to learn reasonable performance or necessitate a large amount of data. Our insight is that when humans learn to drive, they will 1) make decisions over the high-level skill space instead of the low-level control space and 2) leverage expert prior knowledge rather than learning from scratch. Inspired by this, we propose ASAP-RL, an efficient reinforcement learning algorithm for autonomous driving that simultaneously leverages motion skills and expert priors. We first parameterized motion skills, which are diverse enough to cover various complex driving scenarios and situations. A skill parameter inverse recovery method is proposed to convert expert demonstrations from control space to skill space. A simple but effective double initialization technique is proposed to leverage expert priors while bypassing the issue of expert suboptimality and early performance degradation. We validate our proposed method on interactive dense-traffic driving tasks given simple and sparse rewards. Experimental results show that our method can lead to higher learning efficiency and better driving performance relative to previous methods that exploit skills and priors differently. Code is open-sourced to facilitate further research.

Explanation is a key component for the adoption of reinforcement learning (RL) in many real-world decision-making problems. In the literature, the explanation is often provided by saliency attribution to the features of the RL agent's state. In this work, we propose a complementary approach to these explanations, particularly for offline RL, where we attribute the policy decisions of a trained RL agent to the trajectories encountered by it during training. To do so, we encode trajectories in offline training data individually as well as collectively (encoding a set of trajectories). We then attribute policy decisions to a set of trajectories in this encoded space by estimating the sensitivity of the decision with respect to that set. Further, we demonstrate the effectiveness of the proposed approach in terms of quality of attributions as well as practical scalability in diverse environments that involve both discrete and continuous state and action spaces such as grid-worlds, video games (Atari) and continuous control (MuJoCo). We also conduct a human study on a simple navigation task to observe how their understanding of the task compares with data attributed for a trained RL policy. Keywords -- Explainable AI, Verifiability of AI Decisions, Explainable RL.

Deep Reinforcement Learning (DRL) has exhibited efficacy in resolving the Local Path Planning (LPP) problem. However, such application in the real world is immensely limited due to the deficient efficiency and generalization capability of DRL. To alleviate these two issues, a solution named Color is proposed, which consists of an Actor-Sharer-Learner (ASL) training framework and a mobile robot-oriented simulator Sparrow. Specifically, the ASL framework, intending to improve the efficiency of the DRL algorithm, employs a Vectorized Data Collection (VDC) mode to expedite data acquisition, decouples the data collection from model optimization by multithreading, and partially connects the two procedures by harnessing a Time Feedback Mechanism (TFM) to evade data underuse or overuse. Meanwhile, the Sparrow simulator utilizes a 2D grid-based world, simplified kinematics, and conversion-free data flow to achieve a lightweight design. The lightness facilitates vectorized diversity, allowing diversified simulation setups across extensive copies of the vectorized environments, resulting in a notable enhancement in the generalization capability of the DRL algorithm being trained. Comprehensive experiments, comprising 57 benchmark video games, 32 simulated and 36 real-world LPP scenarios, have been conducted to corroborate the superiority of our method in terms of efficiency and generalization. The code and the video of the experiments can be accessed on our website.

Policy learning (PL) is a module of a task-oriented dialogue system that trains an agent to make actions in each dialogue turn. Imitating human action is a fundamental problem of PL. However, both supervised learning (SL) and reinforcement learning (RL) frameworks cannot imitate humans well. Training RL models require online interactions with user simulators, while simulating complex human policy is hard. Performances of SL-based models are restricted because of the covariate shift problem. Specifically, a dialogue is a sequential decision-making process where slight differences in current utterances and actions will cause significant differences in subsequent utterances. Therefore, the generalize ability of SL models is restricted because statistical characteristics of training and testing dialogue data gradually become different. This study proposed an offline imitation learning model that learns policy from real dialogue datasets and does not require user simulators. It also utilizes state transition information, which alleviates the influence of the covariate shift problem. We introduced a regularization trick to make our model can be effectively optimized. We investigated the performance of our model on four independent public dialogue datasets. The experimental result showed that our model performed better in the action prediction task.

In recent years, decentralized finance (DeFi) has experienced remarkable growth, with various protocols such as lending protocols and automated market makers (AMMs) emerging. Traditionally, these protocols employ off-chain governance, where token holders vote to modify parameters. However, manual parameter adjustment, often conducted by the protocol's core team, is vulnerable to collusion, compromising the integrity and security of the system. Furthermore, purely deterministic, algorithm-based approaches may expose the protocol to novel exploits and attacks. In this paper, we present "Auto.gov", a learning-based on-chain governance framework for DeFi that enhances security and reduces susceptibility to attacks. Our model leverages a deep Q- network (DQN) reinforcement learning approach to propose semi-automated, intuitive governance proposals with quantitative justifications. This methodology enables the system to efficiently adapt to and mitigate the negative impact of malicious behaviors, such as price oracle attacks, more effectively than benchmark models. Our evaluation demonstrates that Auto.gov offers a more reactive, objective, efficient, and resilient solution compared to existing manual processes, thereby significantly bolstering the security and, ultimately, enhancing the profitability of DeFi protocols.