Artificial Intelligence By Example

Q-Learning – MDP 

By adding an MDP Q-Learning decision function (see Chapter 1, Become an Adaptive Thinker) in this CNN-CRLMM program, we are now entering the family of DQNs, such as Google's DeepMind programs. 

In this real-time system, there is no beginning, no end, and no real memory beyond a few frames. 

The MDP.py MDP program has been incorporated into MDP_Graph.py. Thus, only the changes made are described in this chapter.

MDP parameters come right after the import packages. Each vertex of the graph has its own letter and its own location, as shown in the following code snippet:

 38:L=['A','B','C','D','E','F']

When the program runs, the following graph will be displayed with red (target vertices) for this frame. It is then up to the ...