Although deep RL models have shown a great potential for solving various types of tasks with minimal supervision, several key challenges remain in terms of learning rapidly from limited experience, adapting to environmental changes, and generalizing learning from a single task. Recent evidence in decision neuroscience has shown that the human brain has an innate capacity to resolve these issues, leading to optimism regarding the development of neuroscience-inspired solutions toward sample-efficient, adaptive, and generalizable RL algorithms. We show that the computational model, adaptively combining model-based and model-free control, which we term the prefrontal RL, reliably encodes the information of highlevel policy that humans learned, and this model can generalize the learned policy to a wide range of tasks. First, we trained the prefrontal RL, deep RL, and meta RL algorithms on 82 human subjects' data, collected while human participants were performing two-stage Markov decision tasks, in which we experimentally manipulated the goal, state-transition uncertainty, and state-space complexity. In the reliability test, which is based on a combination of the latent behavior profile and the parameter recoverability test, we showed that the prefrontal RL reliably learned the latent policies of the human subjects, while all the other models failed to pass this test. Second, to empirically test the ability to generalize what these models learned from the original task, we situated them in the context of environmental volatility. Specifically, we ran large-scale simulations with 10 different Markov decision tasks, in which latent context variables change over time. Our information-theoretic analysis showed that the prefrontal RL showed the highest level of adaptability and episodic encoding efficacy. To the best of our knowledge, this is the first attempt to formally test the possibility that computational models mimicking the way the brain solves general problems can lead to practical solutions to key challenges in machine learning.

We present a Reverse Reinforcement Learning (Reverse RL) approach for representing retrospective knowledge. General Value Functions (GVFs) have enjoyed great success in representing predictive knowledge, i.e., answering questions about possible future outcomes such as "how much fuel will be consumed in expectation if we drive from A to B?". GVFs, however, cannot answer questions like "how much fuel do we expect a car to have given it is at B at time $t$?". To answer this question, we need to know when that car had a full tank and how that car came to B. Since such questions emphasize the influence of possible past events on the present, we refer to their answers as retrospective knowledge. In this paper, we show how to represent retrospective knowledge with Reverse GVFs, which are trained via Reverse RL. We demonstrate empirically the utility of Reverse GVFs in both representation learning and anomaly detection.

Smart FinTech has emerged as a new area that synthesizes and transforms AI and finance, and broadly data science, machine learning, economics, etc. Smart FinTech also transforms and drives new economic and financial businesses, services and systems, and plays an increasingly important role in economy, technology and society transformation. This article presents a highly summarized research overview of smart FinTech, including FinTech businesses and challenges, various FinTech-associated data and repositories, FinTech-driven business decision and optimization, areas in smart FinTech, and research methods and techniques for smart FinTech.

In this work, we present a learning-based approach to analysis cyberspace configuration. Unlike prior methods, our approach has the ability to learn from past experience and improve over time. In particular, as we train over a greater number of agents as attackers, our method becomes better at rapidly finding attack paths for previously hidden paths, especially in multiple domain cyberspace. To achieve these results, we pose finding attack paths as a Reinforcement Learning (RL) problem and train an agent to find multiple domain attack paths. To enable our RL policy to find more hidden attack paths, we ground representation introduction an multiple domain action select module in RL. By designing a simulated cyberspace experimental environment to verify our method. Our objective is to find more hidden attack paths, to analysis the weakness of cyberspace configuration. The experimental results show that our method can find more hidden multiple domain attack paths than existing baselines methods.

This paper proposes PuRL - a deep reinforcement learning (RL) based algorithm for pruning neural networks. Unlike current RL based model compression approaches where feedback is given only at the end of each episode to the agent, PuRL provides rewards at every pruning step. This enables PuRL to achieve sparsity and accuracy comparable to current state-of-the-art methods, while having a much shorter training cycle. PuRL achieves more than 80% sparsity on the ResNet-50 model while retaining a Top-1 accuracy of 75.37% on the ImageNet dataset. Through our experiments we show that PuRL is also able to sparsify already efficient architectures like MobileNet-V2. In addition to performance characterisation experiments, we also provide a discussion and analysis of the various RL design choices that went into the tuning of the Markov Decision Process underlying PuRL. Lastly, we point out that PuRL is simple to use and can be easily adapted for various architectures.

In this work, we propose KeRNS: an algorithm for episodic reinforcement learning in non-stationary Markov Decision Processes (MDPs) whose state-action set is endowed with a metric. Using a non-parametric model of the MDP built with time-dependent kernels, we prove a regret bound that scales with the covering dimension of the state-action space and the total variation of the MDP with time, which quantifies its level of non-stationarity. Our method generalizes previous approaches based on sliding windows and exponential discounting used to handle changing environments. We further propose a practical implementation of KeRNS, we analyze its regret and validate it experimentally.

In this paper, we establish a theoretical comparison between the asymptotic mean-squared error of Double Q-learning and Q-learning. Our result builds upon an analysis for linear stochastic approximation based on Lyapunov equations and applies to both tabular setting and with linear function approximation, provided that the optimal policy is unique and the algorithms converge. We show that the asymptotic mean-squared error of Double Q-learning is exactly equal to that of Q-learning if Double Q-learning uses twice the learning rate of Q-learning and outputs the average of its two estimators. We also present some practical implications of this theoretical observation using simulations.

Learning to plan for long horizons is a central challenge in episodic reinforcement learning problems. A fundamental question is to understand how the difficulty of the problem scales as the horizon increases. Here the natural measure of sample complexity is a normalized one: we are interested in the number of episodes it takes to provably discover a policy whose value is $\varepsilon$ near to that of the optimal value, where the value is measured by the normalized cumulative reward in each episode. In a COLT 2018 open problem, Jiang and Agarwal conjectured that, for tabular, episodic reinforcement learning problems, there exists a sample complexity lower bound which exhibits a polynomial dependence on the horizon -- a conjecture which is consistent with all known sample complexity upper bounds. This work refutes this conjecture, proving that tabular, episodic reinforcement learning is possible with a sample complexity that scales only logarithmically with the planning horizon. In other words, when the values are appropriately normalized (to lie in the unit interval), this results shows that long horizon RL is no more difficult than short horizon RL, at least in a minimax sense. Our analysis introduces two ideas: (i) the construction of an $\varepsilon$-net for optimal policies whose log-covering number scales only logarithmically with the planning horizon, and (ii) the Online Trajectory Synthesis algorithm, which adaptively evaluates all policies in a given policy class using sample complexity that scales with the log-covering number of the given policy class. Both may be of independent interest.

The use of experience plays a key role in reinforcement learning (RL). How best to use this data is one of the central problems of this field. As RL agents have advanced over recent years, taking on bigger and more complex problems (Atari, Go, StarCraft, Dota), the generated data has grown in both size and complexity. To cope with this complexity many RL systems split the learning problem into two distinct parts: experience producers (actors) and experience consumers (learners) — allowing these different parts to run in parallel. Often a data storage system lies at the intersection between these two components. The question of how to efficiently store and transport the data is itself a challenging engineering problem.

This work is devoted to unresolved problems of Artificial General Intelligence - the inefficiency of transfer learning. One of the mechanisms that are used to solve this problem in the area of reinforcement learning is a model-based approach. In the paper we are expanding the schema networks method which allows to extract the logical relationships between objects and actions from the environment data. We present algorithms for training a Delta Schema Network (DSN), predicting future states of the environment and planning actions that will lead to positive reward. DSN shows strong performance of transfer learning on the classic Atari game environment.