Metadata-Version: 2.1
Name: svrg-optimizer-keras
Version: 0.1.0
Summary: SVGR optimizer for Keras
Home-page: https://github.com/tilkb/SVRGoptimizerKeras
Author: Bence Tilk
Author-email: bence.tilk@gmail.com
License: UNKNOWN
Keywords: SVRG,Keras,Optimizer
Platform: UNKNOWN
Classifier: Programming Language :: Python :: 3
Classifier: License :: OSI Approved :: MIT License
Classifier: Operating System :: OS Independent
Description-Content-Type: text/markdown
Requires-Dist: keras (>=2.2)

# Stochastic Variance Reduction Gradient Descent (SVRG) optimizer for Keras

## Reference
https://papers.nips.cc/paper/4937-accelerating-stochastic-gradient-descent-using-predictive-variance-reduction.pdf


## Test results on MNIST
The following figure shows the train loss.
![MNIST](MNIST.png)
__________________________________
The results on test set:
|        |SVRG     |SGD       |
|--------|---------|----------|
|Loss    |0.0893619|0.05765046|
|Accuracy|0.9722   |0.9817    |




