24th LSI Design Contests-in Okinawa  Design Specification - 4-3

4-3.Learning results

The Q value table after actual learning is shown below.

Table 2 : Q-value table after learning

Learning result table

Referring to Table 2, the route taken when an agent selects an action is shown in Fig. 3 below.

Result maze5-5

Fig 3 : Agent's path

Figure 3 shows that they are learning how to get money without going through the devil trout.

The following is a pdf summarizing Q-learning.

pdf file:Q-learning.pdf
<<Back                 Next>>