Publications

We introduce illusory attacks, a novel form of adversarial attacks on reinforcement learning agents that is of bounded statistical detectability.

We propose DEXTER, a novel out-of-distribution dynamics detection method achieving state-of-the-art in this setting.

We propose a perfectly-secure steganography algorithm for arbitrary covertext distributions.

Our approach, Randomized Entity-wise Factorization for Imagined Learning (REFIL) answers the question “expected utility of each agent when only considering a randomly selected sub-group of its observed entities?”.

We propose QMIX, a novel value-based method that can train decentralised policies in a centralised end-to-end fashion.

We propose Model-Free Opponent Shaping (M-FOS) to address the issue of shaping the learning process of their opponents without require long-horizon gradients.

We introduce a novel theoretical framework, named Mirror Learning, which provides theoretical guarantees to a large class of algorithms, including TRPO and PPO.

We propose a belief learning paradigm that can maintain beliefs over rollouts of policies not seen at training time, and can thus decode and adapt to novel conventions at test time.

We propose a perfectly-secure steganography algorithm for arbitrary covertext distributions.

In this paper we develop a new and efficient amortized importance sampling estimator for rejection sampling.

In this paper we explore the Mirror Learning space by meta-learning a “drift” function.

We propose a novel framework and algorithm for cheap talk discovery (CTD) and cheap talk utilization (CTU).

We present a novel equivariant network architecture for use in Dec-POMDPs that prevents the agent from learning policies which break symmetries, doing so more effectively than prior methods.

we suggest a novel reinforcement learning setting that can be used to efficiently generate arbitrary adversarial perturbations using deep multi-agent reinforcement learning.

This thesis presents a number of significant contributions deep multi-agent reinforcement for cooperative control within the framework of centralised training with decentralised execution and two associated novel benchmark suites.

We introduce RainBench, a new multi-modal benchmark dataset for data-driven precipitation forecasting.

We used deep multi-agent reinforcement learning to solve an energy systems wargame wherein players simulate IOC decision-making.

We demonstrate that, despite its various theoretical shortcomings, Independent PPO (IPPO), a form of independent learning in which each agent simply estimates its local value function, can perform just as well as or better than state-of-the-art joint learning approaches.

This paper introduces Multi-Agent Mujoco, an easily extensible multi-agent benchmark suite for robotic control in continuous action spaces

We used deep multi-agent reinforcement learning to solve an energy systems wargame wherein players simulate IOC decision-making.

We propose multi-agent common knowledge reinforcement learning (MACKRL).

We propose multi-agent common knowledge reinforcement learning (MACKRL).

We suggest treating Stratospheric Aersol Injection (SAI) as a high-dimensional control problem, with policies trained according to a context-sensitive reward function within the Deep Reinforcement Learning (DRL) paradigm.

We propose the StarCraft Multi-Agent Challenge (SMAC) to measure real progress in MARL.

We introduce an approach that allows one to treat a large class of population-based epidemiology simulators as probabilistic generative models.

We present the first safe removal bound for data points whichdoes not rely on spectral properties of the kernel matrix.

We prove that the ZX-calculus is incomplete for quantum mechanics.

A physically-motivated genetic algorithm (GA) and full enumeration for a tile-based model of self-assembly (JaTAM) is implemented using a graphics processing unit (GPU).