A Novel Reinforcement Learning Architecture for Continuous State and Action Spaces

Figure 5

Learning curves of the SARSA A²C algorithm using different numbers of multilayer perceptrons.