Multi-class Classification on Volkert
0.3237LossRL with rejection-based reward
Evaluation results
| Method | Links | |
|---|---|---|
| RL with rejection-based rewardArchitecture=64-128-48-16, Number of Parameters=27,0502022.04 | 0.3237 | |
| reference architectureArchitecture=48-160-32-144, Number of Parameters=27,8822022.04 | 0.3244 | |
| RL with rejection-based rewardArchitecture=48-160-32-144, Number of Parameters=27,8822022.04 | 0.3244 | |
| RL with Abs RewardArchitecture=80-48-112-32, Number of Parameters=27,8022022.04 | 0.3274 | |
| RL with Abs RewardArchitecture=96-64-32-48, Number of Parameters=27,7382022.04 | 0.3302 | |
| RL with Abs RewardArchitecture=96-80-16-48, Number of Parameters=27,7382022.04 | 0.3302 | |
| RL with Abs RewardArchitecture=96-48-32-96, Number of Parameters=27,7382022.04 | 0.3305 | |
| RENAArchitecture=32-48-24-512, Number of Parameters=26,4822022.04 | 0.3389 | |
| MNasNet (reward type 1)Architecture=32-32-224-24, Number of Parameters=19,8902022.04 | 0.3392 | |
| MNasNet (reward type 2)Architecture=80-24-24-24, Number of Parameters=17,8742022.04 | 0.3521 |