Both genetic algorithms (GAs) and temporal difference (TD) methods have proven effective at solving reinforcement learning (RL) problems. However, since few rigorous empirical comparisons have been conducted, there are no general guidelines describing the methods ’ relative strengths and weaknesses. This paper presents the results of a detailed empirical comparison between a GA and a TD method in Keepaway, a standard RL benchmark domain based on robot soccer. In particular, we compare the performance of NEAT [19], a GA that evolves neural networks, with Sarsa [16, 17], a popular TD method. The results demonstrate that NEAT can learn better policies in this task, though it requires more evaluations to do so. Additional experiments in two variations of Keepaway demonstrate that Sarsa learns better policies when the task is fully observable and NEAT learns faster when the task is deterministic. Together, these results help isolate the factors critical to the performance of each method and yield insights into their general strengths and weaknesses.

Autonomic Computing is a concept that brings together many fields of computing with the purpose of creating computing systems that self-manage. In its early days it was criticised as being a “hype topic” or a rebadging of some Multi Agent Systems work. In this survey, we hope to show that this was not indeed ’hype ’ and that, though it draws on much work already carried out by the Computer Science and Control communities, its innovation is strong and lies in its robust application to the specific self-management of computing systems. To this end, we first provide an introduction to the motivation and concepts of autonomic computing and describe some research that has been seen as seminal in influencing a large proportion of early work. Taking the components of an established reference model in turn, we discuss the works that have provided significant contributions to that area. We then look at larger scaled systems that compose autonomic systems illustrating the hierarchical nature of their architectures. Autonomicity is not a well defined subject and as such different systems adhere to different degrees of Autonomicity, therefore we cross-slice the body of work in terms of these degrees. From this we list the key applications of autonomic computing and discuss the research work that is missing and what we believe the community should be considering.

Computational learning theory studies mathematical models that allow one to formally analyze and compare the performance of supervised-learning algorithms such as their sample complexity. While existing models such as PAC (Probably Approximately Correct) have played an influential role in understanding the nature of supervised learning, they have not been as successful in reinforcement learning (RL). Here, the fundamental barrier is the need for active exploration in sequential decision problems. An RL agent tries to maximize long-term utility by exploiting its knowledge about the problem, but this knowledge has to be acquired by the agent itself through exploring the problem that may reduce short-term utility. The need for active exploration is common in many problems in daily life, engineering, and sciences. For example, a Backgammon program strives to take good moves to maximize the probability of winning a game, but sometimes it may try novel and possibly harmful moves to discover how the opponent reacts in the hope of discovering a better game-playing strategy. It has been known since the early days of RL that a good tradeoff between exploration and exploitation is critical for the agent to learn fast (i.e., to reach near-optimal strategies

The main objective of transfer in reinforcement learning is to reduce the complexity of learning the solution of a target task by effectively reusing the knowledge retained from solving a set of source tasks. In this paper, we introduce a novel algorithm that transfers samples (i.e., tuples 〈s, a, s ′ , r〉) from source to target tasks. Under the assumption that tasks have similar transition models and reward functions, we propose a method to select samples from the source tasks that are mostly similar to the target task, and, then, to use them as input for batch reinforcementlearning algorithms. As a result, the number of samples an agent needs to collect from the target task to learn its solution is reduced. We empirically show that, following the proposed approach, the transfer of samples is effective in reducing the learning complexity, even when some source tasks are significantly different from the target task. 1.

Improving the sample efficiency of reinforcement learning algorithms to scale up to larger and more realistic domains is a current research challenge in machine learning. Model-based methods use experiential data more efficiently than modelfree approaches but often require exhaustive exploration to learn an accurate model of the domain. We present an algorithm, Reinforcement Learning with Decision Trees (rl-dt), that uses supervised learning techniques to learn the model by generalizing the relative effect of actions across states. Specifically, rl-dt uses decision trees to model the relative effects of actions in the domain. The agent explores the environment exhaustively in early episodes when its model is inaccurate. Once it believes it has developed an accurate model, it exploits its model, taking the optimal action at each step. The combination of the learning approach with the targeted exploration policy enables fast learning of the model. The sample efficiency of the algorithm is evaluated empirically in comparison to five other algorithms across three domains. rl-dt consistently accrues high cumulative rewards in comparison with the other algorithms tested. Categories andSubjectDescriptors

Uncertainty arises in reinforcement learning from various sources, and therefore it is necessary to consider statistics based on several roll-outs for evaluating behavioral policies. We add an adaptive uncertainty handling based on Hoeffding and empirical Bernstein races to the CMA-ES, a variable metric evolution strategy proposed for direct policy search. The uncertainty handling adjusts individually the number of episodes considered for the evaluation of a policy. The performance estimation is kept just accurate enough for a sufficiently good ranking of candidate policies, which is in turn sufficient for the CMA-ES to find better solutions. This increases the learning speed as well as the robustness of the algorithm. 1.

In this paper, we address the task of mapping high-level instructions to sequences of commands in an external environment. Processing these instructions is challenging—they posit goals to be achieved without specifying the steps required to complete them. We describe a method that fills in missing information using an automatically derived environment model that encodes states, transitions, and commands that cause these transitions to happen. We present an efficient approximate approach for learning this environment model as part of a policygradient reinforcement learning algorithm for text interpretation. This design enables learning for mapping high-level instructions, which previous statistical methods cannot handle. 1 1

Abstract. The covariance matrix adaptation evolution strategy (CMA-ES) is suggested for solving problems described by Markov decision processes. The algorithm is compared with a state-of-the-art policy gradient method and stochastic search on the double cart-pole balancing task using linear policies. The CMA-ES proves to be much more robust than the gradient-based approach in this scenario. 1

On behalf of:

In several agent-oriented scenarios in the real world, an autonomous agent that is situated in an unknown environment must learn through a process of trial and error to take actions that result in long-term benefit. Reinforcement Learning (or sequential decision making) is a paradigm well-suited to this requirement. Value function-based methods and policy search methods are contrasting approaches to solve reinforcement learning tasks. While both classes of methods benefit from independent theoretical analyses, these often fail to extend to the practical situations in which the methods are deployed. We conduct an empirical study to examine the strengths and weaknesses of these approaches by introducing a suite of test domains that can be varied for problem size, stochasticity, function approximation, and partial observability. Our results indicate clear patterns in the domain characteristics for which each class of methods excels. We investigate whether their strengths can be combined, and develop an approach to achieve that purpose. The effectiveness of this approach is also demonstrated on the challenging benchmark task of robot soccer Keepaway. We highlight several lines of inquiry that emanate from this study.