Meta-learning methods leverage past experience to learn data-driven inductive biases from related problems, increasing learning efficiency on new tasks. This ability renders them particularly suitable for sequential decision making with limited experience. Within this problem family, we argue for the use of such approaches in the study of model-based approaches to Bayesian Optimisation, contextual bandits and Reinforcement Learning. We approach the problem by learning distributions over functions using Neural Processes (NPs), a recently introduced probabilistic meta-learning method. This allows the treatment of model uncertainty to tackle the exploration/exploitation dilemma. We show that NPs are suitable for sequential decision making on a diverse set of domains, including adversarial task search, recommender systems and model-based reinforcement learning.

Off-Policy reinforcement learning (RL) is an important class of methods for many problem domains, such as robotics, where the cost of collecting data is high and on-policy methods are consequently intractable. Standard methods for applying Q-learning to continuous-valued action domains involve iteratively sampling the Q-function to find a good action (e.g. via hill-climbing), or by learning a policy network at the same time as the Q-function (e.g. DDPG). Both approaches make tradeoffs between stability, speed, and accuracy. We propose a novel approach, called Cross-Entropy Guided Policies, or CGP, that draws inspiration from both classes of techniques. CGP aims to combine the stability and performance of iterative sampling policies with the low computational cost of a policy network. Our approach trains the Q-function using iterative sampling with the Cross-Entropy Method (CEM), while training a policy network to imitate CEM's sampling behavior. We demonstrate that our method is more stable to train than state of the art policy network methods, while preserving equivalent inference time compute costs, and achieving competitive total reward on standard benchmarks.

Humans tend to learn complex abstract concepts faster if examples are presented in a structured manner. For instance, when learning how to play a board game, usually one of the first concepts learned is how the game ends, i.e. the actions that lead to a terminal state (win, lose or draw). The advantage of learning end-games first is that once the actions which lead to a terminal state are understood, it becomes possible to incrementally learn the consequences of actions that are further away from a terminal state - we call this an end-game-first curriculum. Currently the state-of-the-art machine learning player for general board games, AlphaZero by Google DeepMind, does not employ a structured training curriculum; instead learning from the entire game at all times. By employing an end-game-first training curriculum to train an AlphaZero inspired player, we empirically show that the rate of learning of an artificial player can be improved during the early stages of training when compared to a player not using a training curriculum.

We propose a new objective, the counterfactual objective, unifying existing objectives for off-policy policy gradient algorithms in the continuing reinforcement learning (RL) setting. Compared to the commonly used excursion objective, which can be misleading about the performance of the target policy when deployed, our new objective better predicts such performance. We prove the Generalized Off-Policy Policy Gradient Theorem to compute the policy gradient of the counterfactual objective and use an emphatic approach to get an unbiased sample from this policy gradient, yielding the Generalized Off-Policy Actor-Critic (Geoff-PAC) algorithm. We demonstrate the merits of Geoff-PAC over existing algorithms in Mujoco robot simulation tasks, the first empirical success of emphatic algorithms in prevailing deep RL benchmarks.

Reinforcement learning (RL) is a widely used approach for controlling systems with unknown or time-varying dynamics. Even though RL does not require a model of the system, it is known to be faster and safer when using models learned online. We propose to employ symbolic regression (SR) to construct parsimonious process models described by analytic equations for real-time RL control. We have tested our method with two different state-of-the-art SR algorithms which automatically search for equations that fit the measured data. In addition to the standard problem formulation in the state-space domain, we show how the method can also be applied to input-output models of the NARX (nonlinear autoregressive with exogenous input) type. We present the approach on three simulated examples with up to 14-dimensional state space: an inverted pendulum, a mobile robot, and a biped walking robot. A comparison with deep neural networks and local linear regression shows that SR in most cases outperforms these commonly used alternative methods. We demonstrate on a real pendulum system that the analytic model found enables RL to successfully perform the swing-up task, based on a model constructed from only 100 data samples.

The challenge in controlling stochastic systems in which random events can set the system on catastrophic trajectories is to develop a robust ability to respond to such events without significantly compromising the optimality of the baseline control policy. Drug resistance can emerge from random and variable mutations in targeted cell populations; in the absence of an appropriate dosing policy, emergent resistant subpopulations can proliferate and lead to treatment failure. Dynamic feedback dosage control holds promise in combatting this phenomenon, but cell population evolutionary dynamics are complex, stochastic, and often high-dimensional, posing significant challenges to system control. This paper presents CelluDose, a deep reinforcement learning closed-loop dynamic control prototype for automated precision drug dosing targeting stochastic and heterogeneous cell proliferation. Developing optimal dosing schedules for preventing therapy-induced drug resistance involves a tradeoff between the effective suppression of emergent resistant cell subpopulations and the use of conservative dosages and a preference for first-line drugs. CelluDose is trained on model simulations of cell population evolutionary dynamics that combine a system of stochastic differential equations and the additional occurrence of random perturbing events. Both the single-drug and combination therapy policies obtained in training exhibit a 100% success rate at suppressing simulated heterogeneous harmful cell growth and responding to diverse system fluctuations and perturbations within the alloted time and using conservative dosing. The policies obtained were found to be highly robust to model parameter changes and fluctuations not introduced during training.

Recently, reinforcement learning (RL) algorithms have demonstrated remarkable success in learning complicated behaviors from minimally processed input. However, most of this success is limited to simulation. While there are promising successes in applying RL algorithms directly on real systems, their performance on more complex systems remains bottle-necked by the relative data inefficiency of RL algorithms. Domain randomization is a promising direction of research that has demonstrated impressive results using RL algorithms to control real robots. At a high level, domain randomization works by training a policy on a distribution of environmental conditions in simulation. If the environments are diverse enough, then the policy trained on this distribution will plausibly generalize to the real world. A human-specified design choice in domain randomization is the form and parameters of the distribution of simulated environments. It is unclear how to the best pick the form and parameters of this distribution and prior work uses hand-tuned distributions. This extended abstract demonstrates that the choice of the distribution plays a major role in the performance of the trained policies in the real world and that the parameter of this distribution can be optimized to maximize the performance of the trained policies in the real world

Reinforcement learning algorithms rely on exploration to discover new behaviors, which is typically achieved by following a stochastic policy. In continuous control tasks, policies with a Gaussian distribution have been widely adopted. Gaussian exploration however does not result in smooth trajectories that generally correspond to safe and rewarding behaviors in practical tasks. In addition, Gaussian policies do not result in an effective exploration of an environment and become increasingly inefficient as the action rate increases. This contributes to a low sample efficiency often observed in learning continuous control tasks. We introduce a family of stationary autoregressive (AR) stochastic processes to facilitate exploration in continuous control domains. We show that proposed processes possess two desirable features: subsequent process observations are temporally coherent with continuously adjustable degree of coherence, and the process stationary distribution is standard normal. We derive an autoregressive policy (ARP) that implements such processes maintaining the standard agent-environment interface. We show how ARPs can be easily used with the existing off-the-shelf learning algorithms. Empirically we demonstrate that using ARPs results in improved exploration and sample efficiency in both simulated and real world domains, and, furthermore, provides smooth exploration trajectories that enable safe operation of robotic hardware.

Various implementations of Deep Reinforcement Learning (RL) demonstrated excellent performance on tasks that can be solved by trained policy, but they are not without drawbacks. Deep RL suffers from high sensitivity to noisy and missing input and adversarial attacks. To mitigate these deficiencies of deep RL solutions, we suggest involving spiking neural networks (SNNs). Previous work has shown that standard Neural Networks trained using supervised learning for image classification can be converted to SNNs with negligible deterioration in performance. In this paper, we convert Q-Learning ReLU-Networks (ReLU-N) trained using reinforcement learning into SNN. We provide a proof of concept for the conversion of ReLU-N to SNN demonstrating improved robustness to occlusion and better generalization than the original ReLU-N. Moreover, we show promising initial results with converting full-scale Deep Q-networks to SNNs, paving the way for future research.

Recently, there have been several high-profile achievements of agents learning to play games against humans and beat them. We consider an alternative approach that instead addresses game design for a better player experience by training human-like game agents. Specifically, we study the problem of training game agents in service of the development processes of the game developers that design, build, and operate modern games. We highlight some of the ways in which we think intelligent agents can assist game developers to understand their games, and even to build them. Our early results using the proposed agent framework mark a few steps toward addressing the unique challenges that game developers face.