Research Article
An Empirical Investigation of Transfer Effects for Reinforcement Learning
Table 3
Detailed training results of nontransfer and transfer methods to solve sorting 7 numbers for 30 episodes.
| nā=ā7 | | NonTrans_Tr_Steps | Trans_Tr_Steps | Ratio_Tr_Steps | NonTrans_Br_Capacity | Trans_Br_Capacity | Ratio_Br_Capacity |
| 0 | 7444 | 3725 | 2.00 | 0.0575 | 0.0476 | 1.21 | 1 | 17430 | 10013 | 1.74 | 0.0895 | 0.0761 | 1.18 | 2 | 10969 | 3716 | 2.95 | 0.0605 | 0.0494 | 1.22 | 3 | 11175 | 2908 | 3.84 | 0.0541 | 0.0420 | 1.29 | 4 | 9032 | 2744 | 3.29 | 0.0514 | 0.0417 | 1.23 | 5 | 3097 | 731 | 4.24 | 0.0257 | 0.0233 | 1.10 | 6 | 16747 | 15702 | 1.07 | 0.0830 | 0.0868 | 0.96 | 7 | 6947 | 4555 | 1.53 | 0.0566 | 0.0524 | 1.08 | 8 | 5964 | 3726 | 1.60 | 0.0502 | 0.0488 | 1.03 | 9 | 4137 | 1214 | 3.41 | 0.0290 | 0.0273 | 1.06 | 10 | 11132 | 12738 | 0.87 | 0.0710 | 0.0782 | 0.91 | 11 | 6983 | 9039 | 0.77 | 0.0594 | 0.0660 | 0.90 | 12 | 7727 | 1751 | 4.41 | 0.0408 | 0.0316 | 1.29 | 13 | 12421 | 28476 | 0.44 | 0.0877 | 0.1083 | 0.81 | 14 | 14832 | 25429 | 0.58 | 0.0920 | 0.1187 | 0.77 | 15 | 12450 | 7392 | 1.68 | 0.0769 | 0.0689 | 1.12 | 16 | 10787 | 6533 | 1.65 | 0.0539 | 0.0529 | 1.02 | 17 | 8659 | 22808 | 0.38 | 0.1045 | 0.1001 | 1.04 | 18 | 7670 | 2634 | 2.91 | 0.0428 | 0.0387 | 1.10 | 19 | 8086 | 9071 | 0.89 | 0.0615 | 0.0659 | 0.93 | 20 | 9687 | 6631 | 1.46 | 0.0556 | 0.0553 | 1.00 | 21 | 2474 | 580 | 4.27 | 0.0314 | 0.0296 | 1.06 | 22 | 10906 | 15964 | 0.68 | 0.0664 | 0.0851 | 0.78 | 23 | 11889 | 5882 | 2.02 | 0.0587 | 0.0514 | 1.14 | 24 | 5962 | 4259 | 1.40 | 0.0478 | 0.0493 | 0.97 | 25 | 19346 | 13054 | 1.48 | 0.0886 | 0.0767 | 1.16 | 26 | 5705 | 3114 | 1.83 | 0.0468 | 0.0396 | 1.18 | 27 | 7431 | 12660 | 0.59 | 0.0642 | 0.0762 | 0.84 | 28 | 3096 | 927 | 3.34 | 0.0249 | 0.0246 | 1.01 | 29 | 4668 | 953 | 4.90 | 0.0337 | 0.0257 | 1.31 |
|
|