Learning, controls, computation
Model-based reinforcement learning
Supervisor: Will Browne
I love the idea of understanding the "limits of intelligence". Our limits, the limits of computers and the limits of the universe. What is computationally/physically possible given a set of resources?
Some problems are easy to solve, yet others are just fundamentally hard (like planning a road trip - travelling salesman problem, see P vs NP). I am curious about what computational (or cognitive, if you like) structures/functions are required to efficiently solve everyday problems and to understand the world around us.
If an agent is given access to a memory buffer, can certain problems be solved more efficiently than others? An answer to this would hint at the advantage an animal would have in evolving (working) memory.