List your company
ResearchFormerly WizWand

Semi-supervised classification on MNIST 100 labels

0.0096Error RateADGM
-0.0003360.0667320.13380.200868Nov 18, 2015Sep 18, 2016Jul 20, 2017May 21, 2018Mar 22, 2019Jan 21, 2020Nov 21, 2020
Updated 8mo ago

Evaluation results

MethodLinks
ADGM2015.110.0096
Ladder Networks2015.110.0106
Adversarial Autoencoderstraining=end-to-end2015.110.019
CatGAN2015.110.0191
VAT2015.110.0233
Domain-VAEBackbone=4-layer MLP, Latent dimension (z)=10, Number of labels=100, Number of parameters=~1M2020.110.027
SHOT-VAEBackbone=4-layer MLP, Latent dimension (z)=10, Number of labels=100, Number of parameters=~1M2020.110.0312
smooth-ELBOBackbone=4-layer MLP, Latent dimension (z)=10, Number of labels=100, Number of parameters=~1M2020.110.0314
VAE (M1 + M2)2015.110.0333
M1+M2Backbone=4-layer MLP, Latent dimension (z)=10, Number of labels=100, Number of parameters=~1M2020.110.0333
Hyperspherical-VAEBackbone=4-layer MLP, Latent dimension (z)=10, Number of labels=100, Number of parameters=~1M2020.110.052
Disentangled-VAEBackbone=4-layer MLP, Latent dimension (z)=10, Number of labels=100, Number of parameters=~1M2020.110.0971
VAE (M1) + TSVM2015.110.1182
VAE (M2)2015.110.1197
M2Backbone=4-layer MLP, Latent dimension (z)=10, Number of labels=100, Number of parameters=~1M2020.110.1197
NN Baseline2015.110.258