Convergence of Synchronous Reinforcement Learning with Linear Function Approximation
Artur Merke - University of Dortmund
Ralf Schoknecht - University of Karlsruhe
Synchronous reinforcement learning (RL) algorithms with linear function approximation are representable as inhomogeneous matrix iterations of a special form (Schoknecht & Merke, 2003). In this paper we state conditions of convergence for general inhomogeneous matrix iterations and prove that they are both necessary and sufficient. This result extends the work presented in (Schoknecht & Merke, 2003), where only a sufficient condition of convergence was proved. As the condition of convergence is necessary and sufficient, the new result is suitable to prove convergence AND divergence of RL algorithms with function approximation. We use the theorem to deduce a new concise proof of convergence for the synchronous residual gradient algorithm (Baird, 1995). Moreover, we derive a counterexample for which the uniform RL algorithm (Merke & Schoknecht, 2002) diverges. This yields a negative answer to the open question if the uniform RL algorithm converges for arbitrary multiple transitions.