Further reading

The asynchronous team model and its applications to fixed points, optimization, dynamic programming, and networks are developed by Bertsekas and Tsitsiklis [ 1989 ] . Tsitsiklis [ 1994 ] combines parallel asynchronous iteration theory with stochastic approximation to prove convergence of Q-learning under a general coordinate-update model. These established results motivate, but do not by themselves prove, the schedule-descent and tangent-transport obligations posed here.