Roland Dubb

Roland is doing his masters at the university of cape town at the department of mathematics and applied mathematics. For his masters he is looking at addressing the problem of algorithm performance evaluation in the field of cooperative multi-agent reinforcement Learning. Roland’s masters work saw him partner with Instadeep Ltd. in the writing of Towards a Standardised Performance Evaluation Protocol for Cooperative MARL. Roland’s honours work explored the relationship between reinforcement learning and active inference. In partnership with the department of statistical sciences he researched how deep reinforcement learning methods can aid in the scaling of active inference agents. In the context of the exploration-exploitation dilemma, he also researched how the mathematical origins of the active inference agent’s exploratory drive might benefit reinforcement learning agents. Reinforcement learning, neuroscientific applications, causality and general intelligence are among Roland’s research interests.

Honours papers: