Dyna-style reinforcement learning is a powerful approach for problems where not much real data is available. The main idea is to supplement real trajectories, or sequences of sampled states over time, with simulated ones sampled from a learned model of the environment. However, in large state spaces, the problem of learning a good generative model of the environment has been open so far. We propose to use deep belief networks to learn an environment model for use in Dyna. We present our approach and validate it empirically on problems where the state observations consist of images. Our results demonstrate that using deep belief networks, which are full generative models, significantly outperforms the use of linear expectation models, proposed in Sutton et al. (2008).

Hierarchical reinforcement learning (HRL) is a promising approach to extend traditional reinforcement learning (RL) methods to solve more complex tasks. Yet, the majority of current HRL methods require careful task-specific design and on-policy training, making them difficult to apply in real-world scenarios. In this paper, we study how we can develop HRL algorithms that are general, in that they do not make onerous additional assumptions beyond standard RL algorithms, and efficient, in the sense that they can be used with modest numbers of interaction samples, making them suitable for real-world problems such as robotic control. For generality, we develop a scheme where lower-level controllers are supervised with goals that are learned and proposed automatically by the higher-level controllers. To address efficiency, we propose to use off-policy experience for both higher and lower-level training. This poses a considerable challenge, since changes to the lower-level behaviors change the action space for the higher-level policy, and we introduce an off-policy correction to remedy this challenge. This allows us to take advantage of recent advances in off-policy model-free RL to learn both higher- and lower-level policies using substantially fewer environment interactions than on-policy algorithms. We term the resulting HRL agent HIRO and find that it is generally applicable and highly sample-efficient. Our experiments show that HIRO can be used to learn highly complex behaviors for simulated robots, such as pushing objects and utilizing them to reach target locations, learning from only a few million samples, equivalent to a few days of real-time interaction. In comparisons with a number of prior HRL methods, we find that our approach substantially outperforms previous state-of-the-art techniques.

Training a task-completion dialogue agent via reinforcement learning (RL) is costly because it requires many interactions with real users. One common alternative is to use a user simulator. However, a user simulator usually lacks the language complexity of human interlocutors and the biases in its design may tend to degrade the agent. To address these issues, we present Deep Dyna-Q, which to our knowledge is the first deep RL framework that integrates planning for task-completion dialogue policy learning. We incorporate into the dialogue agent a model of the environment, referred to as the world model, to mimic real user response and generate simulated experience. During dialogue policy learning, the world model is constantly updated with real user experience to approach real user behavior, and in turn, the dialogue agent is optimized using both real experience and simulated experience. The effectiveness of our approach is demonstrated on a movie-ticket booking task in both simulated and human-in-the-loop settings.

Cooling system plays a key role in modern data center. Developing an optimal control policy for data center cooling system is a challenging task. The prevailing approaches often rely on approximated system models that are built upon the knowledge of mechanical cooling, electrical and thermal management, which is difficult to design and may lead to sub-optimal or unstable performances. In this paper we propose to utilize the large amount of monitoring data in data center to optimize the control policy. To do so, we cast the cooling control policy design into an energy cost minimization problem with temperature constraints, and tab it into the emerging deep reinforcement learning (DRL) framework. Specifically, we propose an end-to-end neural control algorithm that is based on the actor-critic framework and the deep deterministic policy gradient (DDPG) technique. To improve the robustness of the control algorithm, we test various DRL related optimization techniques, such as recurrent decision making, discounted return, different neural network architectures, and different stochastic gradient descent algorithms, and adding additional constraints on the output of the policy network. We evaluate the proposed algorithms on the EnergyPlus simulation platform and on a real data trace collected from the National Super Computing Centre (NSCC) of Singapore. Our results show that the proposed end-to-end cooling control algorithm can achieve about 10% cooling cost saving on the simulation platform compared with a canonical two stage optimization algorithm; and it can achieve about 13.6% cooling energy saving on the NSCC data trace. Furthermore, it shows high accuracy in predicting the temperature of the racks (with mean absolute error 0.1 degree) and can control the temperature of the data center zone close to the predefined threshold with variation lower to 0.2 degree.

Which authors of this paper are endorsers? Disable MathJax (What is MathJax?)

This paper presents the Crossmodal Attentive Skill Learner (CASL), integrated with the recently-introduced Asynchronous Advantage Option-Critic (A2OC) architecture [Harb et al., 2017] to enable hierarchical reinforcement learning across multiple sensory inputs. We provide concrete examples where the approach not only improves performance in a single task, but accelerates transfer to new tasks. We demonstrate the attention mechanism anticipates and identifies useful latent features, while filtering irrelevant sensor modalities during execution. We modify the Arcade Learning Environment [Bellemare et al., 2013] to support audio queries, and conduct evaluations of crossmodal learning in the Atari 2600 game Amidar. Finally, building on the recent work of Babaeizadeh et al. [2017], we open-source a fast hybrid CPU-GPU implementation of CASL.

The complexity of dynamics in AI techniques is already approaching that of complex adaptive systems, thus curtailing the feasibility of formal controllability and reachability analysis in the context of AI safety. It follows that the envisioned instances of Artificial General Intelligence (AGI) will also suffer from challenges of complexity. To tackle such issues, we propose the modeling of deleterious behaviors in AI and AGI as psychological disorders, thereby enabling the employment of psychopathological approaches to analysis and control of misbehaviors. Accordingly, we present a discussion on the feasibility of the psychopathological approaches to AI safety, and propose general directions for research on modeling, diagnosis, and treatment of psychological disorders in AGI.

We present a framework and algorithm for peer-to-peer teaching in cooperative multiagent reinforcement learning. Our algorithm, Learning to Coordinate and Teach Reinforcement (LeCTR), trains advising policies by using students' learning progress as a teaching reward. Agents using LeCTR learn to assume the role of a teacher or student at the appropriate moments, exchanging action advice to accelerate the entire learning process. Our algorithm supports teaching heterogeneous teammates, advising under communication constraints, and learns both what and when to advise. LeCTR is demonstrated to outperform the final performance and rate of learning of prior teaching methods on multiple benchmark domains. To our knowledge, this is the first approach for learning to teach in a multiagent setting.

We study the inverse optimal control problem in social sciences: we aim at learning a user's true cost function from the observed temporal behavior. In contrast to traditional phenomenological works that aim to learn a generative model to fit the behavioral data, we propose a novel variational principle and treat user as a reinforcement learning algorithm, which acts by optimizing his cost function. We first propose a unified KL framework that generalizes existing maximum entropy inverse optimal control methods. We further propose a two-step Wasserstein inverse optimal control framework. In the first step, we compute the optimal measure with a novel mass transport equation. In the second step, we formulate the learning problem as a generative adversarial network. In two real world experiments - recommender systems and social networks, we show that our framework obtains significant performance gains over both existing inverse optimal control methods and point process based generative models.

We consider a team of reinforcement learning agents that concurrently operate in a common environment, and we develop an approach to efficient coordinated exploration that is suitable for problems of practical scale. Our approach builds on seed sampling (Dimakopoulou and Van Roy, 2018) and randomized value function learning (Osband et al., 2016). We demonstrate that, for simple tabular contexts, the approach is competitive with previously proposed tabular model learning methods (Dimakopoulou and Van Roy, 2018). With a higher-dimensional problem and a neural network value function representation, the approach learns quickly with far fewer agents than alternative exploration schemes.