- Teacher - wide_resnet_40_4
alpha = 0.1
Student | w/o KD | w/ KD | ours (noise rate = 5 / 10 / 20 / 30%) |
---|---|---|---|
resnet20 | 68.74 | 68.90 | 69.06 / 69.40 / 69.25 / 68.82 |
resnet32 | 70.09 | 71.22 | 70.57 / 71.25 / 71.26 / 71.05 |
resnet44 | 72.01 | 71.92 | 71.98 / 71.79 / 72.50 / 72.46 |
resnet56 | 72.13 | 72.73 | 73.12 / 72.79 / 72.85 / 72.99 |
alpha = 0.2
Student | w/o KD | w/ KD | ours (noise rate = 5 / 10 / 20 / 30%) |
---|---|---|---|
resnet20 | 68.74 | 69.49 | 69.09 / 68.81 / 69.23 / 68.78 |
resnet32 | 70.09 | 71.14 | 71.55 / 71.79 / 70.87 / 70.86 |
resnet44 | 72.01 | 72.75 | 72.60 / 72.47 / 72.45 / 72.06 |
resnet56 | 72.13 | 73.40 | 73.16 / 73.02 / 73.46 / 72.71 |
- Teacher - wide_resnet_40_4
- Student - resnet56
symmetric noise
Label Corrupt. | Teacher | w/o KD | w/ KD | ours (noise rate = 10 / 20 / 30%) |
---|---|---|---|---|
20% | 70.14 | 64.67 | 66.03 | 65.72 / 65.82 / 66.19 |
40% | 64.40 | 56.90 | 58.93 | 58.55 / 58.27 / 58.97 |
60% | 55.74 | 47.10 | 48.54 | 46.31 / 49.02 / 48.76 |
pairflip
Label Corrupt. | Teacher | w/o KD | w/ KD | ours (noise rate = 10 / 20 / 30%) |
---|---|---|---|---|
20% | 72.01 | 67.94 | 69.21 | 69.16 / 68.80 / 68.80 |
40% | 57.36 | 51.83 | 55.49 | 54.71 / 55.80 / 56.24 |
- Learning From Noisy Labels With Deep Neural Networks: A Survey
- Learning With Instance-Dependent Label Noise: A Sample Sieve Approach