The \textit{Smoothed Bellman Error Embedding} algorithm~\citep{dai2018sbeed}, known as SBEED, was proposed as a provably convergent reinforcement learning algorithm with general nonlinear function approximation. It has been successfully implemented with neural networks and achieved strong empirical results. In this work, we study the theoretical behavior of SBEED in batch-mode reinforcement learning. We prove a near-optimal performance guarantee that depends on the representation power of the used function classes and a tight notion of the distribution shift. Our results improve upon prior guarantees for SBEED in ~\citet{dai2018sbeed} in terms of the dependence on the planning horizon and on the sample size. Our analysis builds on the recent work of ~\citet{Xie2020} which studies a related algorithm MSBO, that could be interpreted as a \textit{non-smooth} counterpart of SBEED.

Formation mechanisms are fundamental to the study of complex networks, but learning them from observations is challenging. In real-world domains, one often has access only to the final constructed graph, instead of the full construction process, and observed graphs exhibit complex structural properties. In this work, we propose GraphOpt, an end-to-end framework that jointly learns an implicit model of graph structure formation and discovers an underlying optimization mechanism in the form of a latent objective function. The learned objective can serve as an explanation for the observed graph properties, thereby lending itself to transfer across different graphs within a domain. GraphOpt poses link formation in graphs as a sequential decision-making process and solves it using maximum entropy inverse reinforcement learning algorithm. Further, it employs a novel continuous latent action space that aids scalability. Empirically, we demonstrate that GraphOpt discovers a latent objective transferable across graphs with different characteristics. GraphOpt also learns a robust stochastic policy that achieves competitive link prediction performance without being explicitly trained on this task and further enables construction of graphs with properties similar to those of the observed graph.

Online policy optimization (OPO) views policy optimization for sequential decision making as an online learning problem. In this framework, the algorithm designer defines a sequence of online loss functions such that the regret rate in online learning implies the policy convergence rate and the minimal loss witnessed by the policy class determines the policy performance bias. This reduction technique has been successfully applied to solving various policy optimization problems, including imitation learning, structured prediction, and system identification. Interestingly, the policy improvement speed observed in practice is usually much faster than existing theory suggests. In this work, we provide an explanation of this fast policy improvement phenomenon. Let $\epsilon$ denote the policy class bias and assume the online loss functions are convex, smooth, and non-negative. We prove that, after $N$ rounds of OPO with stochastic feedback, the policy converges in $\tilde{O}(1/N + \sqrt{\epsilon/N})$ in both expectation and high probability. In other words, we show that adopting a sufficiently expressive policy class in OPO has two benefits: both the convergence rate increases and the performance bias decreases, as the policy class becomes reasonably rich. This new theoretical insight is further verified in an online imitation learning experiment.

Solving sparse reward tasks through exploration is one of the major challenges in deep reinforcement learning, especially in three-dimensional, partially-observable environments. Critically, the algorithm proposed in this article uses a single human demonstration to solve hard-exploration problems. We train an agent on a combination of demonstrations and own experience to solve problems with variable initial conditions. We adapt this idea and integrate it with the proximal policy optimization (PPO). The agent is able to increase its performance and to tackle harder problems by replaying its own past trajectories prioritizing them based on the obtained reward and the maximum value of the trajectory. We compare different variations of this algorithm to behavioral cloning on a set of hard-exploration tasks in the Animal-AI Olympics environment. To the best of our knowledge, learning a task in a three-dimensional environment with comparable difficulty has never been considered before using only one human demonstration.

Failure of mission-critical equipment interrupts production and results in monetary loss. The risk of unplanned equipment downtime can be minimized through Predictive Maintenance of revenue generating assets to ensure optimal performance and safe operation of equipment. However, the increased sensorization of the equipment generates a data deluge, and existing machine-learning based predictive model alone becomes inadequate for timely equipment condition predictions. In this paper, a model-free Deep Reinforcement Learning algorithm is proposed for predictive equipment maintenance from an equipment-based sensor network context. Within each equipment, a sensor device aggregates raw sensor data, and the equipment health status is analyzed for anomalous events. Unlike traditional black-box regression models, the proposed algorithm self-learns an optimal maintenance policy and provides actionable recommendation for each equipment. Our experimental results demonstrate the potential for broader range of equipment maintenance applications as an automatic learning framework.

Many dynamic processes, including common scenarios in robotic control and reinforcement learning (RL), involve a set of interacting subprocesses. Though the subprocesses are not independent, their interactions are often sparse, and the dynamics at any given time step can often be decomposed into locally independent causal mechanisms. Such local causal structures can be leveraged to improve the sample efficiency of sequence prediction and off-policy reinforcement learning. We formalize this by introducing local causal models (LCMs), which are induced from a global causal model by conditioning on a subset of the state space. We propose an approach to inferring these structures given an object-oriented state representation, as well as a novel algorithm for model-free Counterfactual Data Augmentation (CoDA). CoDA uses local structures and an experience replay to generate counterfactual experiences that are causally valid in the global model. We find that CoDA significantly improves the performance of RL agents in locally factored tasks, including the batch-constrained and goal-conditioned settings.

In Hierarchical Control, compositionality, abstraction, and task-transfer are crucial for designing versatile algorithms which can solve a variety of problems with maximal representational reuse. We propose a novel hierarchical and compositional framework called Jump-Operator Dynamic Programming for quickly computing solutions within a super-exponential space of sequential sub-goal tasks with ordering constraints, while also providing a fast linearly-solvable algorithm as an implementation. This approach involves controlling over an ensemble of reusable goal-conditioned polices functioning as temporally extended actions, and utilizes transition operators called feasibility functions, which are used to summarize initial-to-final state dynamics of the polices. Consequently, the added complexity of grounding a high-level task space onto a larger ambient state-space can be mitigated by optimizing in a lower-dimensional subspace defined by the grounding, substantially improving the scalability of the algorithm while effecting transferable solutions. We then identify classes of objective functions on this subspace whose solutions are invariant to the grounding, resulting in optimal zero-shot transfer.

Reinforcement learning (RL) has been very successful in recent years but, limited by its sample inefficiency, often requires large computational resources. While new methods are being investigated to increase the efficiency of RL algorithms it is critical to enable training at scale, yet using a code-base flexible enough to allow for method experimentation. Here, we present NAPPO, a pytorch-based library for RL which provides scalable proximal policy optimization (PPO) implementations in a simple, modular package. We validate it by replicating previous results on Mujoco and Atari environments. Furthermore, we provide insights on how a variety of distributed training schemes with synchronous and asynchronous communication patterns perform. Finally we showcase NAPPO by obtaining the highest to-date test performance on the Obstacle Tower Unity3D challenge environment. The full source code is available.

What goals should a multi-goal reinforcement learning agent pursue during training in long-horizon tasks? When the desired (test time) goal distribution is too distant to offer a useful learning signal, we argue that the agent should not pursue unobtainable goals. Instead, it should set its own intrinsic goals that maximize the entropy of the historical achieved goal distribution. We propose to optimize this objective by having the agent pursue past achieved goals in sparsely explored areas of the goal space, which focuses exploration on the frontier of the achievable goal set. We show that our strategy achieves an order of magnitude better sample efficiency than the prior state of the art on long-horizon multi-goal tasks including maze navigation and block stacking.

Deep model-based Reinforcement Learning (RL) has the potential to substantially improve the sample-efficiency of deep RL. While various challenges have long held it back, a number of papers have recently come out reporting success with deep model-based methods. This is a great development, but the lack of a consistent metric to evaluate such methods makes it difficult to compare various approaches. For example, the common single-task sample-efficiency metric conflates improvements due to model-based learning with various other aspects, such as representation learning, making it difficult to assess true progress on model-based RL. To address this, we introduce an experimental setup to evaluate model-based behavior of RL methods, inspired by work from neuroscience on detecting model-based behavior in humans and animals. Our metric based on this setup, the Local Change Adaptation (LoCA) regret, measures how quickly an RL method adapts to a local change in the environment. Our metric can identify model-based behavior, even if the method uses a poor representation and provides insight in how close a method's behavior is from optimal model-based behavior. We use our setup to evaluate the model-based behavior of MuZero on a variation of the classic Mountain Car task.