Image Classification on CIFAR-10 + CIFAR-100 open-set noise (test)
79.55Accuracytruncated Lq loss
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| truncated Lq lossNoise rate (η)=40%, Training epochs=1002018.05 | 79.55 | — | — | |
| Wang, et al.Noise rate (η)=40%, Training epochs=1002018.05 | 79.28 | — | — | |
| MAENoise rate (η)=40%, Training epochs=1002018.05 | 75.06 | — | — | |
| Lq lossNoise rate (η)=40%, Training epochs=1002018.05 | 71.1 | — | — | |
| ForwardNoise rate (η)=40%, Training epochs=1002018.05 | 64.18 | — | — | |
| CCENoise rate (η)=40%, Training epochs=1002018.05 | 62.92 | — | — | |
| Co-teachingNoise ratio=40%, Evaluation window=Last 10 epochs2019.01 | — | 79.43 | 79.58 | |
| Co-teaching+Noise ratio=40%, Evaluation window=Last 10 epochs2019.01 | — | 79.28 | 79.74 | |
| IterativeNoise ratio=40%, Evaluation window=Last 10 epochs2019.01 | — | 79.28 | — | |
| MentorNetNoise ratio=40%, Evaluation window=Last 10 epochs2019.01 | — | 79.27 | 79.33 | |
| StandardNoise ratio=40%, Evaluation window=Last 10 epochs2019.01 | — | 62.92 | — |