Trainability issues in quantum policy gradients

TL;DR

This paper investigates the challenges of training quantum policy gradients in reinforcement learning, highlighting issues like barren plateaus and gradient explosion, and proposes conditions for trainability with empirical validation.

Contribution

It identifies key factors affecting trainability of quantum policies and provides conditions to ensure trainability with polynomial measurements, supported by empirical results.

Findings

Standard barren plateaus cause small gradients

Gradient explosion depends on basis-state partitioning

Contiguous partitioning enables trainability with polynomial measurements

Abstract

This research explores the trainability of Parameterized Quantum circuit-based policies in Reinforcement Learning, an area that has recently seen a surge in empirical exploration. While some studies suggest improved sample complexity using quantum gradient estimation, the efficient trainability of these policies remains an open question. Our findings reveal significant challenges, including standard Barren Plateaus with exponentially small gradients and gradient explosion. These phenomena depend on the type of basis-state partitioning and mapping these partitions onto actions. For a polynomial number of actions, a trainable window can be ensured with a polynomial number of measurements if a contiguous-like partitioning of basis-states is employed. These results are empirically validated in a multi-armed bandit environment.