The Markov decision process (MDP) formulation used to model many real-world sequential decision making problems does not capture the setting where the set of available decisions (actions) at each time step is stochastic. Recently, the stochastic action set Markov decision process (SAS-MDP) formulation has been proposed, which captures the concept of a stochastic action set. In this paper we argue that existing RL algorithms for SAS-MDPs suffer from divergence issues, and present new algorithms for SAS-MDPs that incorporate variance reduction techniques unique to this setting, and provide conditions for their convergence. We conclude with experiments that demonstrate the practicality of our approaches using several tasks inspired by real-life use cases wherein the action set is stochastic.

Policy gradients methods are perhaps the most widely used class of reinforcement learning algorithms. These methods apply to complex, poorly understood, control problems by performing stochastic gradient descent over a parameterized class of polices. Unfortunately, even for simple control problems solvable by classical techniques, policy gradient algorithms face non-convex optimization problems and are widely understood to converge only to local minima. This work identifies structural properties -- shared by finite MDPs and several classic control problems -- which guarantee that policy gradient objective function has no suboptimal local minima despite being non-convex. When these assumptions are relaxed, our work gives conditions under which any local minimum is near-optimal, where the error bound depends on a notion of the expressive capacity of the policy class.

Autonomous multiple tasks learning is a fundamental capability to develop versatile artificial agents that can act in complex environments. In real-world scenarios, tasks may be interrelated (or "hierarchical") so that a robot has to first learn to achieve some of them to set the preconditions for learning other ones. Even though different strategies have been used in robotics to tackle the acquisition of interrelated tasks, in particular within the developmental robotics framework, autonomous learning in this kind of scenarios is still an open question. Building on previous research in the framework of intrinsically motivated open-ended learning, in this work we describe how this question can be addressed working on the level of task selection, in particular considering the multiple interrelated tasks scenario as an MDP where the system is trying to maximise its competence over all the tasks.

Temporal point process is an expressive tool for modeling event sequences over time. In this paper, we take a reinforcement learning view whereby the observed sequences are assumed to be generated from a mixture of latent policies. The purpose is to cluster the sequences with different temporal patterns into the underlying policies while learning each of the policy model. The flexibility of our model lies in: i) all the components are networks including the policy network for modeling the intensity function of temporal point process; ii) to handle varying-length event sequences, we resort to inverse reinforcement learning by decomposing the observed sequence into states (RNN hidden embedding of history) and actions (time interval to next event) in order to learn the reward function, thus achieving better performance or increasing efficiency compared to existing methods using rewards over the entire sequence such as log-likelihood or Wasserstein distance. We adopt an expectation-maximization framework with the E-step estimating the cluster labels for each sequence, and the M-step aiming to learn the respective policy. Extensive experiments show the efficacy of our method against state-of-the-arts.

Reinforcement learning has traditionally occupied a niche status in the world of artificial intelligence. But reinforcement learning has started to assume a larger role in many AI initiatives in the past few years. Its application sweet spot is in calculation of optimal actions to be taken by agents in environmentally contextualized decision scenarios. Using trial-and-error approaches to maximize an algorithmic reward function, reinforcement learning is well suited to many adaptive-control and multiagent automation applications in IT operations management, energy, health care, commerce, finance, transportation, and finance. And it's being used to train the AI that powers both its traditional focus areas--robotics, gaming, and simulation--and a new generation of AI solutions in edge analytics, natural language processing, machine translation, computer vision, and digital assistants.

Communicating the goal of a task to another person is easy: we can use language, show them an image of the desired outcome, point them to a how-to video, or use some combination of all of these. On the other hand, specifying a task to a robot for reinforcement learning requires substantial effort. Most prior work that has applied deep reinforcement learning to real robots makes uses of specialized sensors to obtain rewards or studies tasks where the robot's internal sensors can be used to measure reward. Since such instrumentation needs to be done for any new task that we may wish to learn, it poses a significant bottleneck to widespread adoption of reinforcement learning for robotics, and precludes the use of these methods directly in open-world environments that lack this instrumentation. We have developed an end-to-end method that allows robots to learn from a modest number of images that depict successful completion of a task, without any manual reward engineering.

Binary determination of the presence of objects is one of the problems where humans perform extraordinarily better than computer vision systems, in terms of both speed and preciseness. One of the possible reasons is that humans can skip most of the clutter and attend only on salient regions. Recurrent attention models (RAM) are the first computational models to imitate the way humans process images via the REINFORCE algorithm. Despite that RAM is originally designed for image recognition, we extend it and present recurrent existence determination, an attention-based mechanism to solve the existence determination. Our algorithm employs a novel $k$-maximum aggregation layer and a new reward mechanism to address the issue of delayed rewards, which would have caused the instability of the training process. The experimental analysis demonstrates significant efficiency and accuracy improvement over existing approaches, on both synthetic and real-world datasets.

In recent years, reinforcement learning has seen incremental growth in replacing classical dynamic programming in the field of control engineering due to it making limited to no assumptions about the dynamics of the system. Instead, it depends upon universal approximating capabilities of the control structure to develop a good control function through trial and error experimentation. The challenge of this approach is to efficiently carry out the exploration, which allows the controller to adapt to a control strategy with satisfactory global performance. We can envision the implausibility of directly employing reinforcement learning approach in designing a controller for a physical system, as the controller may crash during thousands or even tens of thousands of trials needed before it finds a stable control function, thereby making it an impractical practice for designing robust controllers. Since conducting trials, in reality, is often infeasible, usually, a mathematical model of the physical system is constructed in the form of a simulator, the controller is designed for the model, and then the controller is implemented on the physical system. If there are substantial differences between the model and the physical system, often called the reality gap, then the controller may operate with compromised performance and possibly be unstable. Physical systems often possess underlying dynamics that are difficult to measure accurately such as friction, density distribution, and unknown torques. Furthermore, the dynamics of the system often change over time; the change can be gradual such as when devices wear or new systems break-in or the change can be abrupt as in the catastrophic failure of a sub-component or the replacement of an old part with a new one.

In an effort to better understand the different ways in which the discount factor affects the optimization process in reinforcement learning, we designed a set of experiments to study each effect in isolation. Our analysis reveals that the common perception that poor performance of low discount factors is caused by (too) small action-gaps requires revision. We propose an alternative hypothesis, which identifies the size-difference of the action-gap across the state-space as the primary cause. We then introduce a new method that enables more homogeneous action-gaps by mapping value estimates to a logarithmic space. We prove convergence for this method under standard assumptions and demonstrate empirically that it indeed enables lower discount factors for approximate reinforcement-learning methods. This in turn allows tackling a class of reinforcement-learning problems that are challenging to solve with traditional methods.

Off-policy reinforcement learning aims to leverage experience collected from prior policies for sample-efficient learning. However, in practice, commonly used off-policy approximate dynamic programming methods based on Q-learning and actor-critic methods are highly sensitive to the data distribution, and can make only limited progress without collecting additional on-policy data. As a step towards more robust off-policy algorithms, we study the setting where the off-policy experience is fixed and there is no further interaction with the environment. We identify bootstrapping error as a key source of instability in current methods. Bootstrapping error is due to bootstrapping from actions that lie outside of the training data distribution, and it accumulates via the Bellman backup operator. We theoretically analyze bootstrapping error, and demonstrate how carefully constraining action selection in the backup can mitigate it. Based on our analysis, we propose a practical algorithm, bootstrapping error accumulation reduction (BEAR). We demonstrate that BEAR is able to learn robustly from different off-policy distributions, including random and suboptimal demonstrations, on a range of continuous control tasks.