Meta-reinforcement learning (meta-RL) algorithms allow for agents to learn new behaviors from small amounts of experience, mitigating the sample inefficiency problem in RL. However, while meta-RL agents can adapt quickly to new tasks at test time after experiencing only a few trajectories, the meta-training process is still sample-inefficient. Prior works have found that in the multi-task RL setting, relabeling past transitions and thus sharing experience among tasks can improve sample efficiency and asymptotic performance. We apply this idea to the meta-RL setting and devise a new relabeling method called Hindsight Foresight Relabeling (HFR). We construct a relabeling distribution using the combination of hindsight, which is used to relabel trajectories using reward functions from the training task distribution, and foresight, which takes the relabeled trajectories and computes the utility of each trajectory for each task. HFR is easy to implement and readily compatible with existing meta-RL algorithms. We find that HFR improves performance when compared to other relabeling methods on a variety of meta-RL tasks. Deep Reinforcement Learning (RL) has achieved success on a wide variety of tasks, ranging from computer games to robotics. However, RL agents are typically trained on a single task and are extremely sample-inefficient, often requiring millions of samples to learn a good policy for just that one task. Ideally, RL agents should be able to utilize their prior knowledge and adapt to tasks quickly, just as humans do.

Over the past decades, researchers have been pushing the limits of Deep Reinforcement Learning (DRL). Although DRL has attracted substantial interest from practitioners, many are blocked by having to search through a plethora of available methodologies that are seemingly alike, while others are still building RL agents from scratch based on classical theories. To address the aforementioned gaps in adopting the latest DRL methods, I am particularly interested in testing out if any of the recent technology developed by the leads in the field can be readily applied to a class of optimal trading problems. Unsurprisingly, many prominent breakthroughs in DRL are investigated and tested on strategic games: from AlphaGo to AlphaStar and at about the same time, OpenAI Five. Thus, in this writing, I want to show precisely how to use a DRL library that is initially built for games in a fundamental trading problem; mean reversion. And by introducing a framework that incorporates economically-motivated function properties, I also demonstrate, through the library, a highly-performant and convergent DRL solution to decision-making financial problems in general.

In this paper, methods have been explored to effectively optimise traffic signal control to minimise waiting times and queue lengths, thereby increasing traffic flow. The traffic intersection was first defined as a Markov Decision Process, and a state representation, actions and rewards were chosen. Simulation of Urban MObility (SUMO) was used to simulate an intersection and then compare a Round Robin Scheduler, a Feedback Control mechanism and two Reinforcement Learning techniques - Deep Q Network (DQN) and Advantage Actor-Critic (A2C), as the policy for the traffic signal in the simulation under different scenarios. Finally, the methods were tested on a simulation of a real-world intersection in Bengaluru, India.

Non-prehensile multi-object rearrangement is a robotic task of planning feasible paths and transferring multiple objects to their predefined target poses without grasping. It needs to consider how each object reaches the target and the order of object movement, which significantly deepens the complexity of the problem. To address these challenges, we propose a hierarchical policy to divide and conquer for non-prehensile multi-object rearrangement. In the high-level policy, guided by a designed policy network, the Monte Carlo Tree Search efficiently searches for the optimal rearrangement sequence among multiple objects, which benefits from imitation and reinforcement. In the low-level policy, the robot plans the paths according to the order of path primitives and manipulates the objects to approach the goal poses one by one. We verify through experiments that the proposed method can achieve a higher success rate, fewer steps, and shorter path length compared with the state-of-the-art.

We offer a historical overview of methodologies for quantifying the notion of risk and optimizing risk-aware autonomous systems, with emphasis on risk-averse settings in which safety may be critical. We categorize and present state-of-the-art approaches, and we describe connections between such approaches and ideas from the fields of decision theory, operations research, reinforcement learning, and stochastic control. The first part of the review focuses on model-based risk-averse methods. The second part discusses methods that blend model-based and model-free techniques for the purpose of designing policies with improved adaptive capabilities. We conclude by highlighting areas for future research.

Artificial intelligence (AI) systems are becoming critical components of today's IT landscapes. Their resilience against attacks and other environmental influences needs to be ensured just like for other IT assets. Considering the particular nature of AI, and machine learning (ML) in particular, this paper provides an overview of the emerging field of resilient AI and presents research issues the authors identify as potential future work.

We propose using regularization for Multi-Agent Reinforcement Learning rather than learning explicit cooperative structures called {\em Multi-Agent Regularized Q-learning} (MARQ). Many MARL approaches leverage centralized structures in order to exploit global state information or removing communication constraints when the agents act in a decentralized manner. Instead of learning redundant structures which is removed during agent execution, we propose instead to leverage shared experiences of the agents to regularize the individual policies in order to promote structured exploration. We examine several different approaches to how MARQ can either explicitly or implicitly regularize our policies in a multi-agent setting. MARQ aims to address these limitations in the MARL context through applying regularization constraints which can correct bias in off-policy out-of-distribution agent experiences and promote diverse exploration. Our algorithm is evaluated on several benchmark multi-agent environments and we show that MARQ consistently outperforms several baselines and state-of-the-art algorithms; learning in fewer steps and converging to higher returns.

Reinforcement learning has been shown to perform a range of complex tasks through interaction with an environment or collected leveraging experience. However, many of these approaches presume optimal or near optimal experiences or the presence of a consistent environment. In this work we propose dual, advantage-based behavior policy based on counterfactual regret minimization. We demonstrate the flexibility of this approach and how it can be adapted to online contexts where the environment is available to collect experiences and a variety of other contexts. We demonstrate this new algorithm can outperform several strong baseline models in different contexts based on a range of continuous environments. Additional ablations provide insights into how our dual behavior regularized reinforcement learning approach is designed compared with other plausible modifications and demonstrates its ability to generalize.

This paper introduces Greedy UnMix (GUM) for cooperative multi-agent reinforcement learning (MARL). Greedy UnMix aims to avoid scenarios where MARL methods fail due to overestimation of values as part of the large joint state-action space. It aims to address this through a conservative Q-learning approach through restricting the state-marginal in the dataset to avoid unobserved joint state action spaces, whilst concurrently attempting to unmix or simplify the problem space under the centralized training with decentralized execution paradigm. We demonstrate the adherence to Q-function lower bounds in the Q-learning for MARL scenarios, and demonstrate superior performance to existing Q-learning MARL approaches as well as more general MARL algorithms over a set of benchmark MARL tasks, despite its relative simplicity compared with state-of-the-art approaches.

This course is in an "early bird" release, and we're still updating and adding content to it, please keep in mind before enrolling that the course is not yet complete. "The future is already here – it's just not very evenly distributed." Have you ever wondered how Artificial Intelligence actually works? Do you want to be able to harness the power of neural networks and reinforcement learning to create intelligent agents that can solve tasks with human level complexity? This is the ultimate course online for learning how to use Python to harness the power of Neural Networks to create Artificially Intelligent agents! This course focuses on a practical approach that puts you in the driver's seat to actually build and create intelligent agents, instead of just showing you small toy examples like many other online courses.