People with type 1 diabetes (T1D) lack the ability to produce the insulin their bodies need. As a result, they must continually make decisions about how much insulin to self-administer to adequately control their blood glucose levels. Longitudinal data streams captured from wearables, like continuous glucose monitors, can help these individuals manage their health, but currently the majority of the decision burden remains on the user. To relieve this burden, researchers are working on closed-loop solutions that combine a continuous glucose monitor and an insulin pump with a control algorithm in an `artificial pancreas.' Such systems aim to estimate and deliver the appropriate amount of insulin. Here, we develop reinforcement learning (RL) techniques for automated blood glucose control. Through a series of experiments, we compare the performance of different deep RL approaches to non-RL approaches. We highlight the flexibility of RL approaches, demonstrating how they can adapt to new individuals with little additional data. On over 2.1 million hours of data from 30 simulated patients, our RL approach outperforms baseline control algorithms: leading to a decrease in median glycemic risk of nearly 50% from 8.34 to 4.24 and a decrease in total time hypoglycemic of 99.8%, from 4,610 days to 6. Moreover, these approaches are able to adapt to predictable meal times (decreasing average risk by an additional 24% as meals increase in predictability). This work demonstrates the potential of deep RL to help people with T1D manage their blood glucose levels without requiring expert knowledge. All of our code is publicly available, allowing for replication and extension.

Building Reinforcement Learning (RL) algorithms which are able to adapt to continuously evolving tasks is an open research challenge. One technology that is known to inherently handle such non-stationary input patterns well is Hierarchical Temporal Memory (HTM), a general and biologically plausible computational model for the human neocortex. As the RL paradigm is inspired by human learning, HTM is a natural framework for an RL algorithm supporting non-stationary environments. In this paper, we present HTMRL, the first strictly HTM-based RL algorithm. We empirically and statistically show that HTMRL scales to many states and actions, and demonstrate that HTM's ability for adapting to changing patterns extends to RL. Specifically, HTMRL performs well on a 10-armed bandit after 750 steps, but only needs a third of that to adapt to the bandit suddenly shuffling its arms. HTMRL is the first iteration of a novel RL approach, with the potential of extending to a capable algorithm for Meta-RL.

Deep reinforcement-learning agents have demonstrated great success on various tasks. However, current methods typically suffer from sample complexity problems when learning in high dimensional observation spaces, which limits the application of deep reinforcement-learning agents to complex, uncertain real-world tasks. In this work, we propose and explore Deep Graph Value Network as a promising method to work around this drawback using a message-passing mechanism. The main idea is that the RL agent should be guided by structured non-neural-network algorithms like dynamic programming. According to recent advances in algorithmic alignment, neural networks with structured computation procedures can be trained efficiently. We demonstrate the potential of graph neural network in supporting sample efficient learning by showing that Deep Graph Value Network can outperform unstructured baselines by a large margin with low sample complexity.

Decisions made by machine learning systems have increasing influence on the world, yet it is common for machine learning algorithms to assume that no such influence exists. An example is the use of the i.i.d. assumption in content recommendation. In fact, the (choice of) content displayed can change users' perceptions and preferences, or even drive them away, causing a shift in the distribution of users. We introduce the term auto-induced distributional shift (ADS) to describe the phenomenon of an algorithm causing a change in the distribution of its own inputs. Our goal is to ensure that machine learning systems do not leverage ADS to increase performance when doing so could be undesirable. We demonstrate that changes to the learning algorithm, such as the introduction of meta-learning, can cause hidden incentives for auto-induced distributional shift (HI-ADS) to be revealed. To address this issue, we introduce `unit tests' and a mitigation strategy for HI-ADS, as well as a toy environment for modelling real-world issues with HI-ADS in content recommendation, where we demonstrate that strong meta-learners achieve gains in performance via ADS. We show meta-learning and Q-learning both sometimes fail unit tests, but pass when using our mitigation strategy.

Deep reinforcement learning (DRL) algorithms have successfully been demonstrated on a range of challenging decision making and control tasks. One dominant component of recent deep reinforcement learning algorithms is the target network which mitigates the divergence when learning the Q function. However, target networks can slow down the learning process due to delayed function updates. Another dominant component especially in continuous domains is the policy gradient method which models and optimizes the policy directly. However, when Q functions are approximated with neural networks, their landscapes can be complex and therefore mislead the local gradient. In this work, we propose a self-regularized and self-guided actor-critic method. We introduce a self-regularization term within the TD-error minimization and remove the need for the target network. In addition, we propose a self-guided policy improvement method by combining policy-gradient with zero-order optimization such as the Cross Entropy Method. It helps to search for actions associated with higher Q-values in a broad neighborhood and is robust to local noise in the Q function approximation. These actions help to guide the updates of our actor network. We evaluate our method on the suite of OpenAI gym tasks, achieving or outperforming state of the art in every environment tested.

Recently, we have seen a rapidly growing adoption of Deep Reinforcement Learning (DRL) technologies. Fully achieving the promise of these technologies in practice is, however, extremely difficult. Users have to invest tremendous efforts in building DRL agents, incorporating the agents into various external training environments, and tuning agent implementation/hyper-parameters so that they can reproduce state-of-the-art (SOTA) performance. In this paper, we propose RLzoo, a new DRL library that aims to make it easy to develop and reproduce DRL algorithms. RLzoo has both high-level APIs and low-level APIs, useful for constructing and customising DRL agents, respectively. It has an adaptive agent construction algorithm that can automatically integrate custom RLzoo agents into various external training environments. To help reproduce the results of SOTA algorithms, RLzoo provides rich reference DRL algorithm implementations and effective hyper-parameter settings. Extensive evaluation results show that RLzoo not only outperforms existing DRL libraries in its simplicity of API design; but also provides the largest number of reference DRL algorithm implementations.

Recent progress in Game AI has demonstrated that given enough data from human gameplay, or experience gained via simulations, machines can rival or surpass the most skilled human players in classic games such as Go, or commercial computer games such as Starcraft. We review the current state-of-the-art through the lens of wargaming, and ask firstly what features of wargames distinguish them from the usual AI testbeds, and secondly which recent AI advances are best suited to address these wargame-specific features.

This paper reviews the current state of the art in Artificial Intelligence (AI) technologies and applications in the context of the creative industries. A brief background of AI, and specifically Machine Learning (ML) algorithms, is provided including Convolutional Neural Network (CNNs), Generative Adversarial Networks (GANs), Recurrent Neural Networks (RNNs) and Deep Reinforcement Learning (DRL). We categorise creative applications into five groups related to how AI technologies are used: i) content creation, ii) information analysis, iii) content enhancement and post production workflows, iv) information extraction and enhancement, and v) data compression. We critically examine the successes and limitations of this rapidly advancing technology in each of these areas. We further differentiate between the use of AI as a creative tool and its potential as a creator in its own right. We foresee that, in the near future, machine learning-based AI will be adopted widely as a tool or collaborative assistant for creativity. In contrast, we observe that the successes of machine learning in domains with fewer constraints, where AI is the `creator', remain modest. The potential of AI (or its developers) to win awards for its original creations in competition with human creatives is also limited, based on contemporary technologies. We therefore conclude that, in the context of creative industries, maximum benefit from AI will be derived where its focus is human centric -- where it is designed to augment, rather than replace, human creativity.

Motivated by high-stakes decision-making domains like personalized medicine where user information is inherently sensitive, we design privacy preserving exploration policies for episodic reinforcement learning (RL). We first provide a meaningful privacy formulation using the notion of joint differential privacy (JDP)--a strong variant of differential privacy for settings where each user receives their own sets of output (e.g., policy recommendations). We then develop a private optimism-based learning algorithm that simultaneously achieves strong PAC and regret bounds, and enjoys a JDP guarantee. Our algorithm only pays for a moderate privacy cost on exploration: in comparison to the non-private bounds, the privacy parameter only appears in lower-order terms. Finally, we present lower bounds on sample complexity and regret for reinforcement learning subject to JDP.

In analogy to compressed sensing, which allows sample-efficient signal reconstruction given prior knowledge of its sparsity in frequency domain, we propose to utilize policy simplicity (Occam's Razor) as a prior to enable sample-efficient imitation learning. We first demonstrated the feasibility of this scheme on linear case where state-value function can be sampled directly. We also extended the scheme to scenarios where only actions are visible and scenarios where the policy is obtained from nonlinear network. The method is benchmarked against behavior cloning and results in significantly higher scores with limited expert demonstrations.