Code Repository for the Paper Ensemble Kalman Filter optimizing Deep Neural Networks: An alternative approach to non-performing Gradient Descent

Description

The Ensemble Kalman Filter(EnKF) can be used as an alternative optimizer when training neural networks, especially in cases where gradient information is not available or backpropagation not applicable.

Figure 1 of the manuscript. It depicts the test error of a Convolutional Neural Network on the MNIST dataset optimized by Stochastic Gradient Descent and Ensemble Kalman Filter. The shaded area shows the standard deviations of ten different runs. Each dot is the test error done on a test set independent of the training set.

Prerequisites

To run the experiments please see code/README.md

Citation

If you use the code or data in your research or just need the citation please cite the work as:

@inproceedings{yegenoglu2020ensemble,
  title={Ensemble Kalman Filter Optimizing Deep Neural Networks: An Alternative Approach to Non-performing Gradient Descent},
  author={Yegenoglu, Alper and Krajsek, Kai and Pier, Sandra Diaz and Herty, Michael},
  booktitle={International Conference on Machine Learning, Optimization, and Data Science},
  pages={78--92},
  year={2020},
  organization={Springer}
}