Image Classification on MNIST uniform label noise η=0.1 (7-fold CV)
97.74AccuracyrSDNet
Evaluation results
| Method | Links | |
|---|---|---|
| rSDNetbeta=0.1, lambda=-0.72026.03 | 97.74 | |
| rSDNetbeta=0, lambda=-0.72026.03 | 97.72 | |
| GCEq=0.52026.03 | 97.66 | |
| GCEq=0.72026.03 | 97.58 | |
| rSDNetbeta=0, lambda=-0.82026.03 | 97.57 | |
| rSDNetbeta=0.1, lambda=-12026.03 | 97.53 | |
| MAELoss=Mean Absolute Error2026.03 | 97.5 | |
| rSDNetbeta=0.05, lambda=-12026.03 | 97.42 | |
| TCCEtruncation_param=0.12026.03 | 95.85 | |
| TCCEtruncation_param=0.22026.03 | 95 | |
| rKLD2026.03 | 90.31 | |
| SCEalpha=0.5, beta=12026.03 | 88.56 | |
| FCLmu=0.252026.03 | 87.96 | |
| TCCEtruncation_param=0.32026.03 | 87.64 | |
| FCLmu=02026.03 | 87.62 | |
| CCELoss=Cross-Entropy2026.03 | 86.91 | |
| FCLmu=0.502026.03 | 14.37 | |
| FCLmu=0.752026.03 | 10.95 |