Delve into the world of reinforcement learning algorithms and apply them to different use-cases via Python. This book covers important topics such as policy gradients and Q learning and utilizes frameworks such as Tensorflow, Keras, and OpenAI Gym. Applied Reinforcement Learning with Python introduces you to the theory behind reinforcement learning (RL) algorithms and the code that will be used to implement them. You will take a guided tour through the features of OpenAI Gym, from utilizing standard libraries to creating your own environments, then discover how to frame reinforcement learning problems so you can research, develop, and deploy RL-based solutions.

The goal is to capture information in a market's order books and use that information to predict market movement/direction. That prediction can enable repricing of orders and more efficient market making. Such an approach allows the market maker to provide liquidity whilst making profits at the same time. Market makers are essential to modern markets. They provide the markets with necessary liquidity and make sure the bid/ask spread is reasonably narrow to allow efficient purchasing.

With the advent of 5G and the research into beyond 5G (B5G) networks, a novel and very relevant research issue is how to manage the coexistence of different types of traffic, each with very stringent but completely different requirements. In this paper we propose a deep reinforcement learning (DRL) algorithm to slice the available physical layer resources between ultra-reliable low-latency communications (URLLC) and enhanced Mobile Broad-Band (eMBB) traffic. Specifically, in our setting the time-frequency resource grid is fully occupied by eMBB traffic and we train the DRL agent to employ proximal policy optimization (PPO), a state-of-the-art DRL algorithm, to dynamically allocate the incoming URLLC traffic by puncturing eMBB codewords. Assuming that each eMBB codeword can tolerate a certain limited amount of puncturing beyond which is in outage, we show that the policy devised by the DRL agent never violates the latency requirement of URLLC traffic and, at the same time, manages to keep the number of eMBB codewords in outage at minimum levels, when compared to other state-of-the-art schemes.

We present a novel off-policy loss function for learning a transition model in model-based reinforcement learning. Notably, our loss is derived from the off-policy policy evaluation objective with an emphasis on correcting distribution shift. Compared to previous model-based techniques, our approach allows for greater robustness under model misspecification or distribution shift induced by learning/evaluating policies that are distinct from the data-generating policy. We provide a theoretical analysis and show empirical improvements over existing model-based off-policy evaluation methods. We provide further analysis showing our loss can be used for off-policy optimization (OPO) and demonstrate its integration with more recent improvements in OPO.

Learning to autonomously navigate the web is a difficult sequential decision making task. The state and action spaces are large and combinatorial in nature, and websites are dynamic environments consisting of several pages. One of the bottlenecks of training web navigation agents is providing a learnable curriculum of training environments that can cover the large variety of real-world websites. Therefore, we propose using Adversarial Environment Generation (AEG) to generate challenging web environments in which to train reinforcement learning (RL) agents. We provide a new benchmarking environment, gMiniWoB, which enables an RL adversary to use compositional primitives to learn to generate arbitrarily complex websites. To train the adversary, we propose a new technique for maximizing regret using the difference in the scores obtained by a pair of navigator agents. Our results show that our approach significantly outperforms prior methods for minimax regret AEG. The regret objective trains the adversary to design a curriculum of environments that are "just-the-right-challenge" for the navigator agents; our results show that over time, the adversary learns to generate increasingly complex web navigation tasks. The navigator agents trained with our technique learn to complete challenging, high-dimensional web navigation tasks, such as form filling, booking a flight etc. We show that the navigator agent trained with our proposed Flexible b-PAIRED technique significantly outperforms competitive automatic curriculum generation baselines -- including a state-of-the-art RL web navigation approach -- on a set of challenging unseen test environments, and achieves more than 80% success rate on some tasks.

Recently, a team of researchers from DeepMind, Google Brain and the University of Toronto unveiled a new reinforcement learning agent known as DreamerV2. This reinforcement learning agent learns behaviours purely from the predictions in the compact latent space of a powerful world model. According to the researchers, DreamerV2 is the first agent to achieve human-level performance on the Atari benchmark. DreamerV2, a collaboration between DeepMind, @GoogleAI and the @UofT, is the first RL agent based on a world model to achieve human-level performance on the Atari benchmark. From driverless cars to beating Go world champions, reinforcement learning has come a long way.

In many learning based control methodologies, learning the unknown dynamic model precedes the control phase, while the aim is to control the system such that it remains in some safe region of the state space. In this work our aim is to guarantee safety while learning and control proceed simultaneously. Specifically, we consider the problem of safe learning in nonlinear control-affine systems subject to unknown additive uncertainty. We model uncertainty as a Gaussian signal and use state measurements to learn its mean and covariance. We provide rigorous time-varying bounds on the mean and covariance of the uncertainty and employ them to modify the control input via an optimisation program with safety constraints encoded as a barrier function on the state space. We show that with an arbitrarily large probability we can guarantee that the state will remain in the safe set, while learning and control are carried out simultaneously, provided that a feasible solution exists for the optimisation problem. We provide a secondary formulation of this optimisation that is computationally more efficient. This is based on tightening the safety constraints to counter the uncertainty about the learned mean and covariance. The magnitude of the tightening can be decreased as our confidence in the learned mean and covariance increases (i.e., as we gather more measurements about the environment). Extensions of the method are provided for Gaussian uncertainties with piecewise constant mean and covariance to accommodate more general environments.

We propose UCBMQ, Upper Confidence Bound Momentum Q-learning, a new algorithm for reinforcement learning in tabular and possibly stage-dependent, episodic Markov decision process. UCBMQ is based on Q-learning where we add a momentum term and rely on the principle of optimism in face of uncertainty to deal with exploration. Our new technical ingredient of UCBMQ is the use of momentum to correct the bias that Q-learning suffers while, at the same time, limiting the impact it has on the second-order term of the regret. For UCBMQ , we are able to guarantee a regret of at most $O(\sqrt{H^3SAT}+ H^4 S A )$ where $H$ is the length of an episode, $S$ the number of states, $A$ the number of actions, $T$ the number of episodes and ignoring terms in poly$log(SAHT)$. Notably, UCBMQ is the first algorithm that simultaneously matches the lower bound of $\Omega(\sqrt{H^3SAT})$ for large enough $T$ and has a second-order term (with respect to the horizon $T$) that scales only linearly with the number of states $S$.

We study the dynamics of temporal-difference learning with neural network-based value function approximation over a general state space, namely, \emph{Neural TD learning}. Existing analysis of neural TD learning relies on either infinite width-analysis or constraining the network parameters in a (random) compact set; as a result, an extra projection step is required at each iteration. This paper establishes a new convergence analysis of neural TD learning \emph{without any projection}. We show that the projection-free TD learning equipped with a two-layer ReLU network of any width exceeding $poly(\overline{\nu},1/\epsilon)$ converges to the true value function with error $\epsilon$ given $poly(\overline{\nu},1/\epsilon)$ iterations or samples, where $\overline{\nu}$ is an upper bound on the RKHS norm of the value function induced by the neural tangent kernel. Our sample complexity and overparameterization bounds are based on a drift analysis of the network parameters as a stopped random process in the lazy training regime.

The OpenSpiel framework provides a collection of environments and algorithm implementations for studying Reinforcement Learning (RL) in games. OpenSpiel includes many popular general-sum, zero-sum, perfect and imperfect information games with episodic interfaces suitable for training RL agents. The algorithms implemented in OpenSpiel are contemporary or state-of-the-art (SOTA) and are designed to be highly configurable and extensible. As stated in the documentation and provided example code, the given default parameters are (in the majority of cases) intended to solve the imperfect information poker variant Kuhn [2]. However, the papers originally proposing many of the OpenSpiel algorithms may not necissarily provide results for this environment and instead report results for more challenging games such as Leduc or Heads up No-Limit Texas Holdem. This limits OpenSpiel users' ability to convinently verify the correctness and performance of algorithim implementations using this tool.