Transfer Learning (TL) has shown great potential to accelerate Reinforcement Learning (RL) by leveraging prior knowledge from past learned policies of relevant tasks. Existing transfer approaches either explicitly computes the similarity between tasks or select appropriate source policies to provide guided explorations for the target task. However, how to directly optimize the target policy by alternatively utilizing knowledge from appropriate source policies without explicitly measuring the similarity is currently missing. In this paper, we propose a novel Policy Transfer Framework (PTF) to accelerate RL by taking advantage of this idea. Our framework learns when and which source policy is the best to reuse for the target policy and when to terminate it by modeling multi-policy transfer as the option learning problem. PTF can be easily combined with existing deep RL approaches. Experimental results show it significantly accelerates the learning process and surpasses state-of-the-art policy transfer methods in terms of learning efficiency and final performance in both discrete and continuous action spaces.

Value estimation is a critical component of the reinforcement learning (RL) paradigm. The question of how to effectively learn predictors for value from data is one of the major problems studied by the RL community, and different approaches exploit structure in the problem domain in different ways. Model learning can make use of the rich transition structure present in sequences of observations, but this approach is usually not sensitive to the reward function. In contrast, model-free methods directly leverage the quantity of interest from the future but have to compose with a potentially weak scalar signal (an estimate of the return). In this paper we develop an approach for representation learning in RL that sits in between these two extremes: we propose to learn what to model in a way that can directly help value prediction. To this end we determine which features of the future trajectory provide useful information to predict the associated return. This provides us with tractable prediction targets that are directly relevant for a task, and can thus accelerate learning of the value function. The idea can be understood as reasoning, in hindsight, about which aspects of the future observations could help past value prediction. We show how this can help dramatically even in simple policy evaluation settings. We then test our approach at scale in challenging domains, including on 57 Atari 2600 games.

From microelectronics to mechanics and machine learning, the modern-day robots are a marvel of multiple engineering disciplines. They use sensors, image processing and reinforcement learning algorithms to move the objects around and move around the obstacles as well. However, this is not the case when it comes to handling objects such as glass. The surface properties of glass are transparent, and non-uniform light reflection makes it difficult for the sensors mounted on the robot to understand how to engage in a simple pick and place operation. To address this problem, researchers at Google AI along with Synthesis AI and Columbia University devised a novel machine-learning algorithm called ClearGrasp, that is capable of estimating accurate 3D data of transparent objects from RGB-D images.

"Generalization" is an AI buzzword these days for good reason: most scientists would love to see the models they're training in simulations and video game environments evolve and expand to take on meaningful real-world challenges -- for example in safety, conservation, medicine, etc. One concerned research area is deep reinforcement learning (DRL), which implements deep learning architectures with reinforcement learning algorithms to enable AI agents to learn the best actions possible to attain their goals in virtual environments. DRL has been widely applied in games and robotics. Such DRL agents have an impressive track record on Starcraft II and Dota-2. But because they were trained in fixed environments, studies suggest DRL agents can fail to generalize to even slight variations of their training environments. In a new paper, researchers from the New York University and Modl.ai, a company applying machine learning to game developing, suggest that simple spacial processing methods such as rotation, translation and cropping could help increase model generality.

Online real-time bidding (RTB) is known as a complex auction game where ad platforms seek to consider various influential key performance indicators (KPIs), like revenue and return on investment (ROI). The trade-off among these competing goals needs to be balanced on a massive scale. To address the problem, we propose a multi-objective reinforcement learning algorithm, named MoTiAC, for the problem of bidding optimization with various goals. Specifically, in MoTiAC, instead of using a fixed and linear combination of multiple objectives, we compute adaptive weights overtime on the basis of how well the current state agrees with the agent's prior. In addition, we provide interesting properties of model updating and further prove that Pareto optimality could be guaranteed. We demonstrate the effectiveness of our method on a real-world commercial dataset. Experiments show that the model outperforms all state-of-the-art baselines.

We develop a generic data-driven method for estimator selection in off-policy policy evaluation settings. We establish a strong performance guarantee for the method, showing that it is competitive with the oracle estimator, up to a constant factor. Via in-depth case studies in contextual bandits and reinforcement learning, we demonstrate the generality and applicability of the method. We also perform comprehensive experiments, demonstrating the empirical efficacy of our approach and comparing with related approaches. In both case studies, our method compares favorably with existing methods.

Automating molecular design using deep reinforcement learning (RL) holds the promise of accelerating the discovery of new chemical compounds. A limitation of existing approaches is that they work with molecular graphs and thus ignore the location of atoms in space, which restricts them to 1) generating single organic molecules and 2) heuristic reward functions. To address this, we present a novel RL formulation for molecular design in Cartesian coordinates, thereby extending the class of molecules that can be built. Our reward function is directly based on fundamental physical properties such as the energy, which we approximate via fast quantum-chemical methods. To enable progress towards de-novo molecular design, we introduce MolGym, an RL environment comprising several challenging molecular design tasks along with baselines. In our experiments, we show that our agent can efficiently learn to solve these tasks from scratch by working in a translation and rotation invariant state-action space.

Gradient-based meta-learners such as Model-Agnostic Meta-Learning (MAML) have shown strong few-shot performance in supervised and reinforcement learning settings. However, specifically in the case of meta-reinforcement learning (meta-RL), we can show that gradient-based meta-learners are sensitive to task distributions. With the wrong curriculum, agents suffer the effects of meta-overfitting, shallow adaptation, and adaptation instability. In this work, we begin by highlighting intriguing failure cases of gradient-based meta-RL and show that task distributions can wildly affect algorithmic outputs, stability, and performance. To address this problem, we leverage insights from recent literature on domain randomization and propose meta Active Domain Randomization (meta-ADR), which learns a curriculum of tasks for gradient-based meta-RL in a similar as ADR does for sim2real transfer. We show that this approach induces more stable policies on a variety of simulated locomotion and navigation tasks. We assess in- and out-of-distribution generalization and find that the learned task distributions, even in an unstructured task space, greatly improve the adaptation performance of MAML. Finally, we motivate the need for better benchmarking in meta-RL that prioritizes \textit{generalization} over single-task adaption performance.

Goal-directed Reinforcement Learning (RL) traditionally considers an agent interacting with an environment, prescribing a real-valued reward to an agent proportional to the completion of some goal. Goal-directed RL has seen large gains in sample efficiency, due to the ease of reusing or generating new experience by proposing goals. In this work, we build on the framework of self-play, allowing an agent to interact with itself in order to make progress on some unknown task. We use Active Domain Randomization and self-play to create a novel, coupled environment-goal curriculum, where agents learn through progressively more difficult tasks and environment variations. Our method, Self-Supervised Active Domain Randomization (SS-ADR), generates a growing curriculum, encouraging the agent to try tasks that are just outside of its current capabilities, while building a domain-randomization curriculum that enables state-of-the-art results on various sim2real transfer tasks. Our results show that a curriculum of co-evolving the environment difficulty along with the difficulty of goals set in each environment provides practical benefits in the goal-directed tasks tested.

Reinforcement learning agents usually learn from scratch, which requires a large number of interactions with the environment. This is quite different from the learning process of human. When faced with a new task, human naturally have the common sense and use the prior knowledge to derive an initial policy and guide the learning process afterwards. Although the prior knowledge may be not fully applicable to the new task, the learning process is significantly sped up since the initial policy ensures a quick-start of learning and intermediate guidance allows to avoid unnecessary exploration. Taking this inspiration, we propose knowledge guided policy network (KoGuN), a novel framework that combines human prior suboptimal knowledge with reinforcement learning. Our framework consists of a fuzzy rule controller to represent human knowledge and a refine module to fine-tune suboptimal prior knowledge. The proposed framework is end-to-end and can be combined with existing policy-based reinforcement learning algorithm. We conduct experiments on both discrete and continuous control tasks. The empirical results show that our approach, which combines human suboptimal knowledge and RL, achieves significant improvement on learning efficiency of flat RL algorithms, even with very low-performance human prior knowledge.