List your company
ResearchFormerly WizWand

Permutation Invariant Image Classification on MNIST (test)

0.57Error RateLadder, full
-0.27685.439111.15516.8709Jul 9, 2015
Updated 8mo ago

Evaluation results

MethodLinks
Ladder, fullNumber of used labels=All2015.070.57
Ladder, only bottom-level costNumber of used labels=All2015.070.59
Virtual AdversarialNumber of used labels=All2015.070.64
AdversarialNumber of used labels=All2015.070.78
Γ-modelNumber of used labels=All2015.070.78
DBM, DropoutNumber of used labels=All2015.070.79
Baseline: MLP, BN, Gaussian noiseNumber of used labels=All2015.070.8
MTCNumber of used labels=All2015.070.81
Ladder, fullNumber of used labels=10002015.070.84
Ladder, only bottom-level costNumber of used labels=10002015.070.9
DGNNumber of used labels=All2015.070.96
Ladder, fullNumber of used labels=1002015.071.06
Ladder, only bottom-level costNumber of used labels=1002015.071.09
AtlasRBFNumber of used labels=All2015.071.31
Virtual AdversarialNumber of used labels=10002015.071.32
Transductive SVMNumber of used labels=All2015.071.4
Semi-sup. EmbeddingNumber of used labels=All2015.071.5
Γ-modelNumber of used labels=10002015.071.53
Virtual AdversarialNumber of used labels=1002015.072.12
DGNNumber of used labels=10002015.072.4
Γ-modelNumber of used labels=1002015.073.06
DGNNumber of used labels=1002015.073.33
Pseudo-labelNumber of used labels=10002015.073.46
MTCNumber of used labels=10002015.073.64
AtlasRBFNumber of used labels=10002015.073.68
Transductive SVMNumber of used labels=10002015.075.38
Baseline: MLP, BN, Gaussian noiseNumber of used labels=10002015.075.7
Semi-sup. EmbeddingNumber of used labels=10002015.075.73
AtlasRBFNumber of used labels=1002015.078.1
Pseudo-labelNumber of used labels=1002015.0710.49
MTCNumber of used labels=1002015.0712.03
Transductive SVMNumber of used labels=1002015.0716.81
Semi-sup. EmbeddingNumber of used labels=1002015.0716.86
Baseline: MLP, BN, Gaussian noiseNumber of used labels=1002015.0721.74