This repository contains the Stochastic Frank Wolfe (SFW) implementation in TensorFlow and Pytorch that was developed alongside the two following publications:
Deep Neural Network Training with Frank-Wolfe (arXiv:2010.07243)
Authors: Sebastian Pokutta, Christoph Spiegel, Max Zimmer
Colab Notebooks to reproduce the exact experiments of the paper:
- Colab Notebook for visualization of constraints (TensorFlow)
- Colab Notebook for sparseness during training (TensorFlow)
- Colab Notebook for comparing stochastic Frank–Wolfe methods (TensorFlow)
- Colab Notebook for large network training (PyTorch)
In case you find the paper or the implementation useful for your own research, please consider citing:
@article{pokutta2020deep,
title={Deep neural network training with frank-wolfe},
author={Pokutta, Sebastian and Spiegel, Christoph and Zimmer, Max},
journal={arXiv preprint arXiv:2010.07243},
year={2020}
}
Projection-Free Adaptive Gradients for Large-Scale Optimization (arXiv:2009.14114)
Authors: Cyrille W. Combettes, Christoph Spiegel, Sebastian Pokutta
Colab Notebooks to reproduce the exact experiments of the paper:
- Colab Notebook for convex objectives (not using this repository)
- Colab Notebook for non-convex objectives (TensorFlow)
In case you find the paper or the implementation useful for your own research, please consider citing:
@article{combettes2020projection,
title={Projection-free adaptive gradients for large-scale optimization},
author={Combettes, Cyrille W and Spiegel, Christoph and Pokutta, Sebastian},
journal={arXiv preprint arXiv:2009.14114},
year={2020}
}