Reproduce CIFAR-10 Semi-Supervised about improved-gan

[christiancosgrove]

After a few more hours of training, test error continues to drop; default hyperparams seem to be working. It seems that training the networks to produce realistic samples and training to do semi-supervised learning happen at different timescales.

[alpapado]

About how many epochs are required before the test error gets close to the value reported in the paper?

[zychen2016]

Hello!why discriminator dense layer just have 10 units? why not 10+1(real/fake)？

[daixiangzi]

Hello!why discriminator dense layer just have 10 units? why not 10+1(real/fake)？

because other(11) class be eliminated,

[daixiangzi]

Has anyone managed to reproduce the exact results for semi-supervised learning using train_cifar_feature_matching.py? With the default hyperparameters and 4000 labeled examples, I'm overfitting and getting 32% test error after 48 epochs. Getting 0.5% training error. Paper claims to get test error of only 18.6% on this task.

Do I need to train longer (for the full 1200 epochs?), or are others having this same problem?

i meet same problem,final your results test erro?