Research Article
An Empirical Investigation of Transfer Effects for Reinforcement Learning
Table 2
Detailed training results of nontransfer and transfer methods to solve sorting 6 numbers for 30 episodes.
| nā=ā6 | | NonTrans_Tr_Steps | Trans_Tr_Steps | Ratio_Tr_Steps | NonTrans_Br_Capacity | Trans_Br_Capacity | Ratio_Br_Capacity |
| 0 | 936 | 417 | 2.24 | 0.0598 | 0.0603 | 0.99 | 1 | 1020 | 508 | 2.01 | 0.0878 | 0.0719 | 1.22 | 2 | 1203 | 684 | 1.76 | 0.1020 | 0.0725 | 1.41 | 3 | 1253 | 411 | 3.05 | 0.0801 | 0.0647 | 1.24 | 4 | 750 | 241 | 3.11 | 0.0620 | 0.0456 | 1.36 | 5 | 1446 | 1344 | 1.08 | 0.1035 | 0.1137 | 0.91 | 6 | 871 | 142 | 6.13 | 0.0612 | 0.0395 | 1.55 | 7 | 1386 | 476 | 2.91 | 0.0878 | 0.0650 | 1.35 | 8 | 972 | 565 | 1.72 | 0.0708 | 0.0717 | 0.99 | 9 | 1272 | 752 | 1.69 | 0.0874 | 0.0746 | 1.17 | 10 | 857 | 426 | 2.01 | 0.0697 | 0.0560 | 1.24 | 11 | 1175 | 3850 | 0.31 | 0.1052 | 0.1300 | 0.81 | 12 | 1199 | 563 | 2.13 | 0.0882 | 0.0673 | 1.31 | 13 | 945 | 543 | 1.74 | 0.0809 | 0.0710 | 1.14 | 14 | 1634 | 915 | 1.79 | 0.1110 | 0.0873 | 1.27 | 15 | 1281 | 944 | 1.36 | 0.0971 | 0.0956 | 1.02 | 16 | 970 | 628 | 1.54 | 0.0847 | 0.0780 | 1.09 | 17 | 1070 | 428 | 2.50 | 0.0719 | 0.0593 | 1.21 | 18 | 3918 | 4929 | 0.79 | 0.1569 | 0.1664 | 0.94 | 19 | 1578 | 955 | 1.65 | 0.1133 | 0.0908 | 1.25 | 20 | 857 | 203 | 4.22 | 0.0530 | 0.0437 | 1.21 | 21 | 1461 | 1008 | 1.45 | 0.1050 | 0.0975 | 1.08 | 22 | 743 | 364 | 2.04 | 0.0639 | 0.0534 | 1.20 | 23 | 1299 | 633 | 2.05 | 0.0866 | 0.0734 | 1.18 | 24 | 1665 | 686 | 2.43 | 0.1037 | 0.0734 | 1.41 | 25 | 4582 | 1216 | 3.77 | 0.1469 | 0.1098 | 1.34 | 26 | 945 | 695 | 1.36 | 0.0768 | 0.0680 | 1.13 | 27 | 4021 | 1201 | 3.35 | 0.1384 | 0.1086 | 1.27 | 28 | 942 | 474 | 1.99 | 0.0737 | 0.0597 | 1.23 | 29 | 1453 | 1276 | 1.14 | 0.1109 | 0.1165 | 0.95 |
|
|