We introduce Wasserstein Policy Optimization (WPO), an actor-critic algorithm for reinforcement learning in continuous action spaces. WPO can be derived as an approximation to Wasserstein gradient flow over the space of all policies projected into a finite-dimensional parameter space (e.g., the weights of a neural network), leading to a simple and completely general closed-form update. The resulting algorithm combines many properties of deterministic and classic policy gradient methods. Like deterministic policy gradients, it exploits knowledge of the *gradient* of the action-value function with respect to the action. Like classic policy gradients, it can be applied to stochastic policies with arbitrary distributions over actions -- without using the reparameterization trick. We show results on the DeepMind Control Suite and a magnetic confinement fusion task which compare favorably with state-of-the-art continuous control methods.

Kernel selection plays a central role in determining the performance of Gaussian Process (GP) models, as the chosen kernel determines both the inductive biases and prior support of functions under the GP prior. This work addresses the challenge of constructing custom kernel functions for high-dimensional GP regression models. Drawing inspiration from recent progress in deep learning, we introduce a novel approach named KITT: Kernel Identification Through Transformers. KITT exploits a transformer-based architecture to generate kernel recommendations in under 0.1 seconds, which is several orders of magnitude faster than conventional kernel search algorithms. We train our model using synthetic data generated from priors over a vocabulary of known kernels. By exploiting the nature of the self-attention mechanism, KITT is able to process datasets with inputs of arbitrary dimension. We demonstrate that kernels chosen by KITT yield strong performance over a diverse collection of regression benchmarks.

Multi-Agent Reinforcement Learning (MARL) considers settings in which a set of coexisting agents interact with one another and their environment. The adaptation and learning of other agents induces non-stationarity in the environment dynamics. This poses a great challenge for value function-based algorithms whose convergence usually relies on the assumption of a stationary environment. Policy search algorithms also struggle in multi-agent settings as the partial observability resulting from an opponent's actions not being known introduces high variance to policy training. Modelling an agent's opponent(s) is often pursued as a means of resolving the issues arising from the coexistence of learning opponents. An opponent model provides an agent with some ability to reason about other agents to aid its own decision making. Most prior works learn an opponent model by assuming the opponent is employing a stationary policy or switching between a set of stationary policies. Such an approach can reduce the variance of training signals for policy search algorithms. However, in the multi-agent setting, agents have an incentive to continually adapt and learn. This means that the assumptions concerning opponent stationarity are unrealistic. In this work, we develop a novel approach to modelling an opponent's learning dynamics which we term Learning to Model Opponent Learning (LeMOL). We show our structured opponent model is more accurate and stable than naive behaviour cloning baselines. We further show that opponent modelling can improve the performance of algorithmic agents in multi-agent settings.

In situations where explicit communication is limited, human collaborators act by learning to: (i) infer meaning behind their partner's actions, and (ii) convey private information about the state to their partner implicitly through actions. The first component of this learning process has been well-studied in multi-agent systems, whereas the second — which is equally crucial for successful collaboration — has not. To mimic both components mentioned above, thereby completing the learning process, we introduce a novel algorithm: Policy Belief Learning (PBL). PBL uses a belief module to model the other agent's private information and a policy module to form a distribution over actions informed by the belief module. Furthermore, to encourage communication by actions, we propose a novel auxiliary reward which incentivizes one agent to help its partner to make correct inferences about its private information. The auxiliary reward for communication is integrated into the learning of the policy module. We evaluate our approach on a set of environments including a matrix game, particle environment and the non-competitive bidding problem from contract bridge. We show empirically that this auxiliary reward is effective and easy to generalize. These results demonstrate that our PBL algorithm can produce strong pairs of agents in collaborative games where explicit communication is disabled.

“Serious games” are becoming extremely relevant to individuals who have specific needs, such as children with an autism spectrum condition (ASC). Often, individuals with an ASC have difficulties in interpreting verbal and nonverbal communication cues during social interactions. The ASC-Inclusion EU-FP7 funded project aims to provide children who have an ASC with a platform to learn emotion expression and recognition, through play in the virtual world. In particular, the ASC-Inclusion platform focuses on the expression of emotion via facial, vocal, and bodily gestures. The platform combines multiple analysis tools, using onboard microphone and webcam capabilities. The platform utilizes these capabilities via training games, text-based communication, animations, video, and audio clips. This paper introduces current findings and evaluations of the ASC-Inclusion platform and provides detailed description for the different modalities.

Online educational tools often generate learning data, and sharing such data between tutors and students can often improve learning outcomes. Unfortunately the process of sharing learning data today is not always transparent to students. Our aim is to improve the transparency and user control aspects of sharing data whilst maintaining the educational utility of data sharing between tutors and students. To do so, we start by surveying the possible methods of sharing data, and we use this to design a token-based scheme for facilitating data sharing. We implemented our scheme and observed it in use by 7,798 students over the course of one year. We find that our proposed scheme provides a good balance between transparency, user control, educational utility and scalability.

We investigate the use of hints as a form of scaffolding for 4,652 eligible users on a large-scale online learning environment called Isaac, which allows users to answer physics questions with up to five hints. We investigate user behaviour when using hints, users' engagement with fading (the process of gradually becoming less reliant on the hints provided), and hint strategies including Decomposition, Correction, Verification, or Comparison. Finally, we present recommendations for the design and development of online teaching tools that provide open access to hints, including a mechanism that may improve the speed at which users begin fading.

We describe a new tool, Equality, for equation entry using free-form layout of components drawn from a palette of symbols. Our approach is designed to enable learners to easily manipulate the structure of their equations, to be functional in both desktop and mobile environments, and to minimize the amount of learning required to use the tool. We present the results of a study comparing a prototype of our approach with Microsoft Equation Editor using a desktop machine. The initial results are promising with participants reporting that the mechanism is easy to learn and an easy way to manipulate their equations. We report the results of the study and the views of the participants and identify how these will inform the future development of Equality.

In this paper we present a novel system for driver-vehicle interaction which combines speech recognition with facial-expression recognition to increase intention recognition accuracy in the presence of engine- and road-noise. Our system would allow drivers to interact with in-car devices such as satellite navigation and other telematic or control systems. We describe a pilot study and experiment in which we tested the system, and show that multimodal fusion of speech and facial expression recognition provides higher accuracy than either would do alone.