Identifying Policy Gradient Subspaces
This incremental finding could lead to more efficient reinforcement learning by improving exploration or enabling second-order optimization.
The paper investigated whether gradients in deep policy gradient methods lie in low-dimensional, slowly-changing subspaces, as seen in supervised learning, and confirmed their existence across various simulated benchmark tasks despite the changing data distribution in reinforcement learning.
Policy gradient methods hold great potential for solving complex continuous control tasks. Still, their training efficiency can be improved by exploiting structure within the optimization problem. Recent work indicates that supervised learning can be accelerated by leveraging the fact that gradients lie in a low-dimensional and slowly-changing subspace. In this paper, we conduct a thorough evaluation of this phenomenon for two popular deep policy gradient methods on various simulated benchmark tasks. Our results demonstrate the existence of such gradient subspaces despite the continuously changing data distribution inherent to reinforcement learning. These findings reveal promising directions for future work on more efficient reinforcement learning, e.g., through improving parameter-space exploration or enabling second-order optimization.