GitHub - s-laine/tempens: Temporal ensembling for semi-supervised learning

Implementation of temporal ensembling and Pi-model. Samuli Laine and Timo Aila, NVIDIA.

Released as part of ICLR 2017 paper submission "Temporal Ensembling for Semi-Supervised Learning".

Additional code (report.py, theano_utils.py, thread_utils.py) by Tero Karras, NVIDIA. Code in zca_bn.py adapted from Tim Salimans' repository at: https://github.com/TimSalimans/weight_norm/blob/master/nn.py

Package versions used when preparing the paper:

To configure a training run, edit config.py. To execute, run train.py.

Provide feedback

Name		Name	Last commit message	Last commit date
Latest commit History 2 Commits
data		data
LICENSE.md		LICENSE.md
README.md		README.md
config.py		config.py
report.py		report.py
theano_utils.py		theano_utils.py
thread_utils.py		thread_utils.py
train.py		train.py
zca_bn.py		zca_bn.py