Classification on CIFAR-10 (test) (r=0.1 to r=0.5)
93.61Accuracy (r=0.1)MLNT
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| MLNTrole=teacher, iteration=3rd2018.12 | 93.61 | 93.25 | 92.82 | 92.3 | 82.09 | |
| MLNTrole=student, iteration=3rd2018.12 | 93.36 | 92.98 | 92.59 | 91.87 | 82.25 | |
| MLNTrole=teacher, iteration=2nd2018.12 | 93.33 | 92.97 | 92.43 | 90.93 | 81.47 | |
| Joint Optimization2018.12 | 93.2 | 92.7 | 92.4 | 91.5 | 84.6 | |
| MLNTrole=teacher, iteration=1st2018.12 | 93.05 | 92.19 | 91.47 | 88.69 | 78.44 | |
| MLNTrole=student, iteration=2nd2018.12 | 93.01 | 92.65 | 91.87 | 90.6 | 81.53 | |
| MLNTrole=student, iteration=1st2018.12 | 92.89 | 91.84 | 90.55 | 88.7 | 79.95 | |
| Forward2018.12 | 92.4 | 91.4 | 91 | 90.3 | 83.8 | |
| CNN-CRF2018.12 | 92 | 91.5 | 90.7 | 89.5 | 84 | |
| Cross Entropyimplementation=original2018.12 | 91.8 | 90.8 | 90 | 87.1 | 77.3 | |
| Cross Entropyimplementation=reproduced2018.12 | 91.04 | 90.19 | 88.88 | 86.34 | 77.48 |