Intelligent agents must pursue their goals in complex environments with partial information and often limited computational capacity. Reinforcement learning methods have achieved great success by creating agents that optimize engineered reward functions, but which often struggle to learn in sparse-reward environments, generally require many environmental interactions to perform well, and are typically computationally very expensive. Active inference is a model-based approach that directs agents to explore uncertain states while adhering to a prior model of their goal behaviour. This paper introduces an active inference agent which minimizes the novel free energy of the expected future. Our model is capable of solving sparse-reward problems with a very high sample efficiency due to its objective function, which encourages directed exploration of uncertain states. Moreover, our model is computationally very light and can operate in a fully online manner while achieving comparable performance to offline RL methods. We showcase the capabilities of our model by solving the mountain car problem, where we demonstrate its superior exploration properties and its robustness to observation noise, which in fact improves performance. We also introduce a novel method for approximating the prior model from the reward function, which simplifies the expression of complex objectives and improves performance over previous active inference approaches.

In the past decade, several studies have explored a new framework for embedding conditional distributions in reproducing kernel Hilbert spaces (RKHS). This approach seeks to represent a conditional distribution as an RKHS element, and thereby reduce the computation of conditional expectations to the evaluation of kernel inner products. Unlike other distribution learning approaches, which often involve density estimation and expensive numerical analysis, the conditional mean embedding (CME) framework exploits the popular kernel trick to allow distributions to be learned directly and efficiently from sample information, and do not require the target distribution to possess a density function. The broad generalizability and computational levity of conditional embeddings have led them to find many applications in reinforcement learning, hypothesis testing, and nonparametric inference [3-5, 17], where conditional relationships are often of pertinent interest. A central issue involved in the conditional embedding framework is the performance of the sample estimator. Despite their successful application, there has been a limited study of optimal learning rates for conditional mean embeddings. Several foundational works [17, 18] established the consistency of the sample embedding estimator, exploring its convergence rate to a "true" embedding in the RKHS norm. These works framed the act of conditioning as a linear operator between two Hilbert spaces, which mapped features of the independent variable in the input space to the mean embeddings of their respective conditional distributions in the output feature space.

Population-based multi-agent reinforcement learning (PB-MARL) refers to the series of methods nested with reinforcement learning (RL) algorithms, which produces a self-generated sequence of tasks arising from the coupled population dynamics. By leveraging auto-curricula to induce a population of distinct emergent strategies, PB-MARL has achieved impressive success in tackling multi-agent tasks. Despite remarkable prior arts of distributed RL frameworks, PB-MARL poses new challenges for parallelizing the training frameworks due to the additional complexity of multiple nested workloads between sampling, training and evaluation involved with heterogeneous policy interactions. To solve these problems, we present MALib, a scalable and efficient computing framework for PB-MARL. Our framework is comprised of three key components: (1) a centralized task dispatching model, which supports the self-generated tasks and scalable training with heterogeneous policy combinations; (2) a programming architecture named Actor-Evaluator-Learner, which achieves high parallelism for both training and sampling, and meets the evaluation requirement of auto-curriculum learning; (3) a higher-level abstraction of MARL training paradigms, which enables efficient code reuse and flexible deployments on different distributed computing paradigms. Experiments on a series of complex tasks such as multi-agent Atari Games show that MALib achieves throughput higher than 40K FPS on a single machine with $32$ CPU cores; 5x speedup than RLlib and at least 3x speedup than OpenSpiel in multi-agent training tasks. MALib is publicly available at https://github.com/sjtu-marl/malib.

Large ride-hailing platforms, such as DiDi, Uber and Lyft, connect tens of thousands of vehicles in a city to millions of ride demands throughout the day, providing great promises for improving transportation efficiency through the tasks of order dispatching and vehicle repositioning. Existing studies, however, usually consider the two tasks in simplified settings that hardly address the complex interactions between the two, the real-time fluctuations between supply and demand, and the necessary coordinations due to the large-scale nature of the problem. In this paper we propose a unified value-based dynamic learning framework (V1D3) for tackling both tasks. At the center of the framework is a globally shared value function that is updated continuously using online experiences generated from real-time platform transactions. To improve the sample-efficiency and the robustness, we further propose a novel periodic ensemble method combining the fast online learning with a large-scale offline training scheme that leverages the abundant historical driver trajectory data. This allows the proposed framework to adapt quickly to the highly dynamic environment, to generalize robustly to recurrent patterns and to drive implicit coordinations among the population of managed vehicles. Extensive experiments based on real-world datasets show considerably improvements over other recently proposed methods on both tasks. Particularly, V1D3 outperforms the first prize winners of both dispatching and repositioning tracks in the KDD Cup 2020 RL competition, achieving state-of-the-art results on improving both total driver income and user experience related metrics.

Counterfactual instances are a powerful tool to obtain valuable insights into automated decision processes, describing the necessary minimal changes in the input space to alter the prediction towards a desired target. Most previous approaches require a separate, computationally expensive optimization procedure per instance, making them impractical for both large amounts of data and high-dimensional data. Moreover, these methods are often restricted to certain subclasses of machine learning models (e.g. differentiable or tree-based models). In this work, we propose a deep reinforcement learning approach that transforms the optimization procedure into an end-to-end learnable process, allowing us to generate batches of counterfactual instances in a single forward pass. Our experiments on real-world data show that our method i) is model-agnostic (does not assume differentiability), relying only on feedback from model predictions; ii) allows for generating target-conditional counterfactual instances; iii) allows for flexible feature range constraints for numerical and categorical attributes, including the immutability of protected features (e.g. gender, race); iv) is easily extended to other data modalities such as images.

We introduce RL-DARTS, one of the first applications of Differentiable Architecture Search (DARTS) in reinforcement learning (RL) to search for convolutional cells, applied to the Procgen benchmark. We outline the initial difficulties of applying neural architecture search techniques in RL, and demonstrate that by simply replacing the image encoder with a DARTS supernet, our search method is sample-efficient, requires minimal extra compute resources, and is also compatible with off-policy and on-policy RL algorithms, needing only minor changes in preexisting code. Surprisingly, we find that the supernet can be used as an actor for inference to generate replay data in standard RL training loops, and thus train end-to-end. Throughout this training process, we show that the supernet gradually learns better cells, leading to alternative architectures which can be highly competitive against manually designed policies, but also verify previous design choices for RL policies.

We present an end-to-end, model-based deep reinforcement learning agent which dynamically attends to relevant parts of its state, in order to plan and to generalize better out-of-distribution. The agent's architecture uses a set representation and a bottleneck mechanism, forcing the number of entities to which the agent attends at each planning step to be small. In experiments with customized MiniGrid environments with different dynamics, we observe that the design allows agents to learn to plan effectively, by attending to the relevant objects, leading to better out-of-distribution generalization.

For continuing environments, reinforcement learning methods commonly maximize a discounted reward criterion with discount factor close to 1 in order to approximate the steady-state reward (the gain). However, such a criterion only considers the long-run performance, ignoring the transient behaviour. In this work, we develop a policy gradient method that optimizes the gain, then the bias (which indicates the transient performance and is important to capably select from policies with equal gain). We derive expressions that enable sampling for the gradient of the bias, and its preconditioning Fisher matrix. We further propose an algorithm that solves the corresponding bi-level optimization using a logarithmic barrier. Experimental results provide insights into the fundamental mechanisms of our proposal.

In this paper, we present a Symbolic Reinforcement Learning (SRL) based architecture for safety control of Radio Access Network (RAN) applications. In particular, we provide a purely automated procedure in which a user can specify high-level logical safety specifications for a given cellular network topology in order for the latter to execute optimal safe performance which is measured through certain Key Performance Indicators (KPIs). The network consists of a set of fixed Base Stations (BS) which are equipped with antennas, which one can control by adjusting their vertical tilt angle. The aforementioned process is called Remote Electrical Tilt (RET) optimization. Recent research has focused on performing this RET optimization by employing Reinforcement Learning (RL) strategies due to the fact that they have self-learning capabilities to adapt in uncertain environments. The term safety refers to particular constraints bounds of the network KPIs in order to guarantee that when the algorithms are deployed in a live network, the performance is maintained. In our proposed architecture the safety is ensured through model-checking techniques over combined discrete system models (automata) that are abstracted through the learning process. We introduce a user interface (UI) developed to help a user set intent specifications to the system, and inspect the difference in agent proposed actions, and those that are allowed and blocked according to the safety specification.

A long-standing challenge in artificial intelligence is lifelong learning. In lifelong learning, many tasks are presented in sequence and learners must efficiently transfer knowledge between tasks while avoiding catastrophic forgetting over long lifetimes. On these problems, policy reuse and other multi-policy reinforcement learning techniques can learn many tasks. However, they can generate many temporary or permanent policies, resulting in memory issues. Consequently, there is a need for lifetime-scalable methods that continually refine a policy library of a pre-defined size. This paper presents a first approach to lifetime-scalable policy reuse. To pre-select the number of policies, a notion of task capacity, the maximal number of tasks that a policy can accurately solve, is proposed. To evaluate lifetime policy reuse using this method, two state-of-the-art single-actor base-learners are compared: 1) a value-based reinforcement learner, Deep Q-Network (DQN) or Deep Recurrent Q-Network (DRQN); and 2) an actor-critic reinforcement learner, Proximal Policy Optimisation (PPO) with or without Long Short-Term Memory layer. By selecting the number of policies based on task capacity, D(R)QN achieves near-optimal performance with 6 policies in a 27-task MDP domain and 9 policies in an 18-task POMDP domain; with fewer policies, catastrophic forgetting and negative transfer are observed. Due to slow, monotonic improvement, PPO requires fewer policies, 1 policy for the 27-task domain and 4 policies for the 18-task domain, but it learns the tasks with lower accuracy than D(R)QN. These findings validate lifetime-scalable policy reuse and suggest using D(R)QN for larger and PPO for smaller library sizes. During their lifetime, animals may be subjected to a large number of unknown tasks. In some environmental conditions, nutritious food sources may be readily available, while in others they may be sparse, hidden, or even poisonous, and dangerous predators may roam in their vicinity. To address these challenging conditions, various behaviours must be selectively combined, such as avoidance, reward-seeking, or even fleeing. When direct perception provides limited or no cues about the current task, animals have to infer the task or use a strategy that works for many different tasks it may encounter. Therefore, a key challenge for learning challenging sequences of tasks is to find a limited number of strategies that work on the large domain of tasks that animals encounter over their lifetime. In artificial intelligence, variants of the above problem have been studied with investigations focusing on two aspects: transfer learning and catastrophic forgetting. Transfer learning is a process in which learners leverage the knowledge gained from a set of previously learned tasks with similar characteristics to a new task, whilst avoiding transferring knowledge that is not relevant [Taylor and Stone, 2009, Pan and Yang, 2010, Lazaric, 2013].