Composition of Web Services Using Markov Decision Processes and Dynamic Programming

Figure 7

Learning times required for a real scenario of Web service composition, plotted in logarithmic scale. Reinforcement learning methods required two orders of magnitude and more time than dynamic programming methods.