Image Classification on noise padded CIFAR-10 (test)
79.12Test Accuracytruncated Lq loss
Evaluation results
| Method | Links | |
|---|---|---|
| truncated Lq lossNoise rate (η)=40%, Training epochs=1002018.05 | 79.12 | |
| Wang, et al.Noise rate (η)=40%, Training epochs=1002018.05 | 78.15 | |
| ForwardNoise rate (η)=40%, Training epochs=1002018.05 | 77.81 | |
| MAENoise rate (η)=40%, Training epochs=1002018.05 | 74.31 | |
| Lq lossNoise rate (η)=40%, Training epochs=1002018.05 | 64.79 | |
| UNICORNN (L=3)#UNITS=128, M=47K2021.03 | 62.4 | |
| CCENoise rate (η)=40%, Training epochs=1002018.05 | 62.38 | |
| CORNN#UNITS=128, M=46K2021.03 | 59 | |
| coRNN# units=128, # params=46k2020.10 | 59 | |
| LIPSCHITZ RNN#UNITS=256, M=158K2021.03 | 55.8 | |
| Lipschitz RNN# units=256, # params=134k2020.10 | 55.2 | |
| GATED ANTI.SYM. RNN#UNITS=256, M=37K2021.03 | 54.7 | |
| Gated anti.sym. RNN# units=256, # params=37k2020.10 | 54.7 | |
| INCREMENTAL RNN#UNITS=128, M=12K2021.03 | 54.5 | |
| Incremental RNN# units=128, # params=12k2020.10 | 54.5 | |
| ANTI.SYM. RNN#UNITS=256, M=36K2021.03 | 48.3 | |
| anti.sym. RNN# units=256, # params=36k2020.10 | 48.3 | |
| FASTRNN#UNITS=128, M=16K2021.03 | 45.8 | |
| FastRNN# units=128, # params=16k2020.10 | 45.8 | |
| LSTM#UNITS=128, M=64K2021.03 | 11.6 | |
| LSTM# units=128, # params=64k2020.10 | 11.6 |