Skip to content

Latest commit

 

History

History
76 lines (44 loc) · 3.26 KB

README.md

File metadata and controls

76 lines (44 loc) · 3.26 KB

A 2-parameter Persistence Layer for Learning

This codebase contains implementation of Generalized Rank Invariant Landscape (GRIL). The accompanying paper can be found at GRIL: A 2-parameter Persistence Based Vectorization for Machine Learning.

Group Information

CGTDA group at Purdue This project is developed by Soham Mukherjee, Cheng Xin, Shreyas N. Samaga and Tamal Dey under the CGTDA research group at Purdue University led by Prof. Tamal Dey.

Acknowledgements

This codebase heavily uses Fast Computation of Zigzag Persistence. The repository for FastZigzag can be found here https://github.com/taohou01/fzz. The software is based on the following paper Fast Computation of Zigzag Persistence.

Instructions

First clone this repo to say $MPML. Then create a conda environment by

conda create -n mpml python=3.9 pytorch=1.12 pyg -c pytorch -c pyg

conda activate mpml

Additional Dependencies:

  1. Boost
  2. OpenMP

Then we need to compile mpml.

cd $MPML
cd gril
python -m pip install -e .

Please follow experiments.ipynb for instructions on how to run the code. You should be able to reproduce the code.

GRIL as topological discriminator

Graph Experiments

You may use run_graph_experiment.sh to reproduce the results in the paper. Please download the precomputed landscapes from this link and unzip the zip file to train the model faster. After unzipping it should have a directory called graph_landscapes.

./run_graph_experiment.sh PROTEINS

Run this script to reproduce the experiment on PROTEINS dataset.

License

THIS SOFTWARE IS PROVIDED "AS-IS". THERE IS NO WARRANTY OF ANY KIND. NEITHER THE AUTHORS NOR PURDUE UNIVERSITY WILL BE LIABLE FOR ANY DAMAGES OF ANY KIND, EVEN IF ADVISED OF SUCH POSSIBILITY.

This software was developed (and is copyrighted by) the CGTDA research group at Purdue University. Please do not redistribute this software. This program is for academic research use only. This software uses the Boost and phat library, which are covered under their own licenses.

Citation

The paper is accepted in ICML TAGML 2023 Workshop.


@InProceedings{pmlr-v221-xin23a,
  title = 	 {GRIL: A $2$-parameter Persistence Based Vectorization for Machine Learning},
  author =       {Xin, Cheng and Mukherjee, Soham and Samaga, Shreyas N. and Dey, Tamal K.},
  booktitle = 	 {Proceedings of 2nd Annual Workshop on Topology, Algebra, and Geometry in Machine Learning (TAG-ML)},
  pages = 	 {313--333},
  year = 	 {2023},
  volume = 	 {221},
  series = 	 {Proceedings of Machine Learning Research},
  month = 	 {28 Jul},
  publisher =    {PMLR},
  pdf = 	 {https://proceedings.mlr.press/v221/xin23a/xin23a.pdf},
  url = 	 {https://proceedings.mlr.press/v221/xin23a.html},
}