General RL algorithms like Q-learning [17], SARSA and TD(λ) [15] have been proved to converge to the globally optimal solution (under certain assumptions) [1][17]. They are very flexible, because they do not require a model of the environment, and have been shown to be effective in solving a variety of RL tasks. This flexibility, however, comes at a certain cost: these RL algorithms require extremely long training to cope with large state space problems. Many different approaches have been proposed for speeding up the RL process. One possible technique is to use function approximation [8], in order to reduce the effect of the "curse of dimensionality".

A technique for speeding up reinforcement learning algorithms by using time manipulation is proposed. It is applicable to failure-avoidance control problems running in a computer simulation. Turning the time of the simulation backwards on failure events is shown to speed up the learning by 260% and improve the state space exploration by 12% on the cart-pole balancing task, compared to the conventional Q-learning and Actor-Critic algorithms.

In this paper, we explore methods for comparing agent behavior with human behavior to assist with validation. Our exploration begins by considering a simple method of behavior comparison. Motivated by shortcomings in this initial approach, we introduce behavior bounding, an automated model-based approach for comparing behavior that is inspired, in part, by Mitchell's Version Spaces. We show that behavior bounding can be used to compactly represent both human and agent behavior. We argue that relatively low amounts of human effort are required to build, maintain, and use the data structures that underlie behavior bounding, and we provide a theoretical basis for these arguments using notions of PAC Learnability. Next, we show empirical results indicating that this approach is effective at identifying differences in certain types of behaviors and that it performs well when compared against our initial benchmark methods. Finally, we demonstrate that behavior bounding can produce information that allows developers to identify and fix problems in an agent's behavior much more efficiently than standard debugging techniques.

We derive an equation for temporal difference learning from statistical principles. Specifically, we start with the variational principle and then bootstrap to produce an updating rule for discounted state value estimates. The resulting equation is similar to the standard equation for temporal difference learning with eligibility traces, so called TD(λ), however it lacks the parameter α that specifies the learning rate. In the place of this free parameter there is now an equation for the learning rate that is specific to each state transition. We experimentally test this new learning rule against TD(λ) and find that it offers superior performance in various settings. Finally, we make some preliminary investigations into how to extend our new temporal difference algorithm to reinforcement learning. To do this we combine our update equation with both Watkins' Q(λ) and Sarsa(λ) and find that it again offers superior performance without a learning rate parameter.

Recent experimental studies have focused on the specialization of different neural structures for different types of instrumental behavior. Recent theoretical work has provided normative accounts for why there should be more than one control system, and how the output of different controllers can be integrated. Two particlar controllers have been identified, one associated with a forward model and the prefrontal cortex and a second associated with computationally simpler, habitual, actor-critic methods and part of the striatum. We argue here for the normative appropriateness of an additional, but so far marginalized control system, associated with episodic memory, and involving the hippocampus and medial temporal cortices. We analyze in depth a class of simple environments to show that episodic control should be useful in a range of cases characterized by complexity and inferential noise, and most particularly at the very early stages of learning, long before habitization has set in. We interpret data on the transfer of control from the hippocampus to the striatum in the light of this hypothesis.

Recently, we have introduced a novel approach to dynamic programming and reinforcement learning that is based on maintaining explicit representations of stationary distributions instead of value functions. In this paper, we investigate the convergence properties of these dual algorithms both theoretically and empirically, and show how they can be scaled up by incorporating function approximation.

Businesses want to save power without sacrificing performance. This paper presents a reinforcement learning approach to simultaneous online management of both performance and power consumption. We apply RL in a realistic laboratory testbed using a Blade cluster and dynamically varying HTTP workload running on a commercial web applications middleware platform. We embed a CPU frequency controller in the Blade servers' firmware, and we train policies for this controller using a multi-criteria reward signal depending on both application performance and CPU power consumption. Our testbed scenario posed a number of challenges to successful use of RL, including multiple disparate reward functions, limited decision sampling rates, and pathologies arising when using multiple sensor readings as state variables. We describe innovative practical solutions to these challenges, and demonstrate clear performance improvements over both hand-designed policies as well as obvious "cookbook" RL implementations.

We study the problem of an apprentice learning to behave in an environment with an unknown reward function by observing the behavior of an expert. We follow on the work of Abbeel and Ng [1] who considered a framework in which the true reward function is assumed to be a linear combination of a set of known and observable features. We give a new algorithm that, like theirs, is guaranteed to learn a policy that is nearly as good as the expert's, given enough examples. However, unlike their algorithm, we show that ours may produce a policy that is substantially better than the expert's. Moreover, our algorithm is computationally faster, is easier to implement, and can be applied even in the absence of an expert. The method is based on a game-theoretic view of the problem, which leads naturally to a direct application of the multiplicative-weights algorithm of Freund and Schapire [2] for playing repeated matrix games. In addition to our formal presentation and analysis of the new algorithm, we sketch how the method can be applied when the transition function itself is unknown, and we provide an experimental demonstration of the algorithm on a toy video-game environment.

Recent experimental studies have focused on the specialization of different neural structures for different types of instrumental behavior. Recent theoretical work has provided normative accounts for why there should be more than one control system, and how the output of different controllers can be integrated. Two particlar controllers have been identified, one associated with a forward model and the prefrontal cortex and a second associated with computationally simpler, habitual, actor-critic methods and part of the striatum. We argue here for the normative appropriateness of an additional, but so far marginalized control system, associated with episodic memory, and involving the hippocampus and medial temporal cortices. We analyze in depth a class of simple environments to show that episodic control should be useful in a range of cases characterized by complexity and inferential noise, and most particularly at the very early stages of learning, long before habitization has set in. We interpret data on the transfer of control from the hippocampus to the striatum in the light of this hypothesis.

Learning in real-world domains often requires to deal with continuous state and action spaces. Although many solutions have been proposed to apply Reinforcement Learning algorithms to continuous state problems, the same techniques can be hardly extended to continuous action spaces, where, besides the computation of a good approximation of the value function, a fast method for the identification of the highest-valued action is needed. In this paper, we propose a novel actor-critic approach in which the policy of the actor is estimated through sequential Monte Carlo methods. The importance sampling step is performed on the basis of the values learned by the critic, while the resampling step modifies the actor's policy. The proposed approach has been empirically compared to other learning algorithms into several domains; in this paper, we report results obtained in a control problem consisting of steering a boat across a river.