Reinforcement learning is the study of decision making over time with consequences. The field has developed systems to make decisions in complex environments based on external, and possibly delayed, feedback. "Microsoft Research, works on developing the theory, algorithms and systems for technology that learns from its own successes (and failures), explores the world "just enough" to learn, and can infer which decisions have led to those outcomes. Our primary goal is reinforcement learning in the real world: understanding how to build systems that work, even when simulation is unavailable and samples are scarce." To celebrate hosting the Reinforcement Learning Israel Meetup organized by the talented Shani Gamrian at the Microsoft Reactor here is a list of three Reinforcement Learning Environments every ML enthusiast should know.

SafeLife is part of a broader PAI initiative to develop benchmarks for safety, fairness, and other ethical objectives for machine learning systems. Since so much of machine learning is driven, shaped, and measured by benchmarks (and the datasets and environments they are based on), we believe it is essential that those benchmarks come to incorporate safety and ethics goals on a widespread basis, and we're working to make that happen.

We seek to align agent behavior with a user's objectives in a reinforcement learning setting with unknown dynamics, an unknown reward function, and unknown unsafe states. The user knows the rewards and unsafe states, but querying the user is expensive. To address this challenge, we propose an algorithm that safely and interactively learns a model of the user's reward function. We start with a generative model of initial states and a forward dynamics model trained on off-policy data. Our method uses these models to synthesize hypothetical behaviors, asks the user to label the behaviors with rewards, and trains a neural network to predict the rewards. The key idea is to actively synthesize the hypothetical behaviors from scratch by maximizing tractable proxies for the value of information, without interacting with the environment. We call this method reward query synthesis via trajectory optimization (ReQueST). We evaluate ReQueST with simulated users on a state-based 2D navigation task and the image-based Car Racing video game. The results show that ReQueST significantly outperforms prior methods in learning reward models that transfer to new environments with different initial state distributions. Moreover, ReQueST safely trains the reward model to detect unsafe states, and corrects reward hacking before deploying the agent.

Most microscopic pedestrian navigation models use the concept of "forces" applied to the pedestrian agents to replicate the navigation environment. While the approach could provide believable results in regular situations, it does not always resemble natural p edestrian navigation behaviour in many typical settings. In our research, we proposed a novel approach using reinforcement learning for simulation of pedestrian agent path planning and collision avoidance problem. The primary focus of this approach is usi ng human perception of the environment and danger awareness of interferences . The implementation of our model has shown that the path planned by the agent shares many similarities with a human pedestrian in several aspects such as following common walking conventions and human behaviours .

Despite the simplicity and intuitive interpretation of Minimum Mean Squared Error (MMSE) estimators, their effectiveness in certain scenarios is questionable. Indeed, minimizing squared errors on average does not provide any form of stability, as the volatility of the estimation error is left unconstrained. When this volatility is statistically significant, the difference between the average and realized performance of the MMSE estimator can be drastically different. To address this issue, we introduce a new risk-aware MMSE formulation which trades between mean performance and risk by explicitly constraining the expected predictive variance of the involved squared error. We show that, under mild moment boundedness conditions, the corresponding risk-aware optimal solution can be evaluated explicitly, and has the form of an appropriately biased nonlinear MMSE estimator. We further illustrate the effectiveness of our approach via several numerical examples, which also showcase the advantages of risk-aware MMSE estimation against risk-neutral MMSE estimation, especially in models involving skewed, heavy-tailed distributions.

We introduce "Search with Amortized Value Estimates" (SAVE), an approach for combining model-free Q-learning with model-based Monte-Carlo Tree Search (MCTS). In SAVE, a learned prior over state-action values is used to guide MCTS, which estimates an improved set of state-action values. The new Q-estimates are then used in combination with real experience to update the prior. This effectively amortizes the value computation performed by MCTS, resulting in a cooperative relationship between model-free learning and model-based search. SAVE can be implemented on top of any Q-learning agent with access to a model, which we demonstrate by incorporating it into agents that perform challenging physical reasoning tasks and Atari. SAVE consistently achieves higher rewards with fewer training steps, and---in contrast to typical model-based search approaches---yields strong performance with very small search budgets. By combining real experience with information computed during search, SAVE demonstrates that it is possible to improve on both the performance of model-free learning and the computational cost of planning.

In this paper we present an end-to-end framework for addressing the problem of dynamic pricing on E-commerce platform using methods based on deep reinforcement learning (DRL). By using four groups of different business data to represent the states of each time period, we model the dynamic pricing problem as a Markov Decision Process (MDP). Compared with the state-of-the-art DRL-based dynamic pricing algorithms, our approaches make the following three contributions. First, we extend the discrete set problem to the continuous price set. Second, instead of using revenue as the reward function directly, we define a new function named difference of revenue conversion rates (DRCR). Third, the cold-start problem of MDP is tackled by pre-training and evaluation using some carefully chosen historical sales data. Our approaches are evaluated by both offline evaluation method using real dataset of Alibaba Inc., and online field experiments on Tmall.com, a major online shopping website owned by Alibaba Inc.. In particular, experiment results suggest that DRCR is a more appropriate reward function than revenue, which is widely used by current literature. In the end, field experiments, which last for months on 1000 stock keeping units (SKUs) of products demonstrate that continuous price sets have better performance than discrete sets and show that our approaches significantly outperformed the manual pricing by operation experts.

We consider the problem of efficient credit assignment in reinforcement learning. In order to efficiently and meaningfully utilize new data, we propose to explicitly assign credit to past decisions based on the likelihood of them having led to the observed outcome. This approach uses new information in hindsight, rather than employing foresight. Somewhat surprisingly, we show that value functions can be rewritten through this lens, yielding a new family of algorithms. We study the properties of these algorithms, and empirically show that they successfully address important credit assignment challenges, through a set of illustrative tasks.

Function approximation is a powerful approach for structuring large decision problems that has facilitated great achievements in the areas of reinforcement learning and game playing. Regression counterfactual regret minimization (RCFR) is a flexible and simple algorithm for approximately solving imperfect information games with policies parameterized by a normalized rectified linear unit (ReLU). In contrast, the more conventional softmax parameterization is standard in the field of reinforcement learning and has a regret bound with a better dependence on the number of actions in the tabular case. We derive approximation error-aware regret bounds for $(\Phi, f)$-regret matching, which applies to a general class of link functions and regret objectives. These bounds recover a tighter bound for RCFR and provides a theoretical justification for RCFR implementations with alternative policy parameterizations ($f$-RCFR), including softmax. We provide exploitability bounds for $f$-RCFR with the polynomial and exponential link functions in zero-sum imperfect information games, and examine empirically how the link function interacts with the severity of the approximation to determine exploitability performance in practice. Although a ReLU parameterized policy is typically the best choice, a softmax parameterization can perform as well or better in settings that require aggressive approximation.

We study reinforcement learning (RL) in a setting with a network of agents whose states and actions interact in a local manner where the objective is to find localized policies such that the (discounted) global reward is maximized. A fundamental challenge in this setting is that the state-action space size scales exponentially in the number of agents, rendering the problem intractable for large networks. In this paper, we propose a Scalable Actor-Critic (SAC) framework that exploits the network structure and finds a localized policy that is a $O(\rho^\kappa)$-approximation of a stationary point of the objective for some $\rho\in(0,1)$, with complexity that scales with the local state-action space size of the largest $\kappa$-hop neighborhood of the network.