Quantile Reinforcement Learning
This work addresses the need for alternative evaluation metrics in reinforcement learning, particularly for episodic problems, but appears incremental as it adapts existing stochastic approximation methods.
The paper tackles the problem of evaluating policies in reinforcement learning by proposing a quantile-based criterion as an alternative to the standard expected sum of rewards, and it demonstrates the algorithm on a model of the TV show 'Who Wants to Be a Millionaire' with unspecified results.
In reinforcement learning, the standard criterion to evaluate policies in a state is the expectation of (discounted) sum of rewards. However, this criterion may not always be suitable, we consider an alternative criterion based on the notion of quantiles. In the case of episodic reinforcement learning problems, we propose an algorithm based on stochastic approximation with two timescales. We evaluate our proposition on a simple model of the TV show, Who wants to be a millionaire.