We introduce a new count-based optimistic exploration algorithm for Reinforcement Learning (RL) that is feasible in environments with high-dimensional state-action spaces. The success of RL algorithms in these domains depends crucially on generalisation from limited training experience. »
AIXI is a universal solution to the reinforcement learning problem; it can learn any computable environment. We formally define death for generally intelligent agents like AIXI, and prove a number of related theorems about their behaviour. »
This thesis investigates two different problems in Reinforcement Learning (RL). The first proposes a new optimistic exploration method for RL that is feasible in very large state spaces. The second part of this work develops an original formalisation of death for RL agents. »
This work has two aims: first, to develop a critique of representational theories of meaning, and second, to outline and defend Merleau-Ponty’s alternative theory of language as gesture. »
