Fig. 5From: Incorporating causal factors into reinforcement learning for dynamic treatment regimes in HIVa Comparison of the performance of direct PG and CPG algorithm; b Dynamic evolution of causal factor C during learningBack to article page