Conditionally Elicitable Dynamic Risk Measures for Deep Reinforcement Learning
This addresses risk management in reinforcement learning for domains like finance, though it appears to be an incremental improvement over existing nested simulation approaches.
The paper tackles risk-sensitive reinforcement learning by proposing a framework to optimize time-consistent dynamic spectral risk measures using conditional elicitability, resulting in a conceptually improved algorithm that avoids nested simulations and demonstrates performance in financial applications like statistical arbitrage and portfolio allocation.
We propose a novel framework to solve risk-sensitive reinforcement learning (RL) problems where the agent optimises time-consistent dynamic spectral risk measures. Based on the notion of conditional elicitability, our methodology constructs (strictly consistent) scoring functions that are used as penalizers in the estimation procedure. Our contribution is threefold: we (i) devise an efficient approach to estimate a class of dynamic spectral risk measures with deep neural networks, (ii) prove that these dynamic spectral risk measures may be approximated to any arbitrary accuracy using deep neural networks, and (iii) develop a risk-sensitive actor-critic algorithm that uses full episodes and does not require any additional nested transitions. We compare our conceptually improved reinforcement learning algorithm with the nested simulation approach and illustrate its performance in two settings: statistical arbitrage and portfolio allocation on both simulated and real data.