Trainability issues in quantum policy gradients
This addresses trainability problems in quantum reinforcement learning for researchers, but it is incremental as it builds on prior empirical work.
The study investigated trainability challenges in quantum policy gradients, revealing issues like Barren Plateaus and gradient explosion that depend on basis-state partitioning, and found that a contiguous-like partitioning can ensure trainability with polynomial measurements for polynomial actions, validated empirically in a multi-armed bandit environment.
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.