BLIS-Net: Classifying and Analyzing Signals on Graphs

BLIS-Net (bi-Lipschitz Scattering Network) is a provably powerful GNN designed for graph signal classification. For further details, please refer to our AISTATS 2024 paper.

Introduction

BLIS-Net consists of four modules/layers:

1. The BLIS-Module
2. Moment aggregation module
3. Embedding/dimensionality reduction layer
4. Classification layer

To accommodate a variety of workflows and tasks, we provide two equivalent implementations of the modules of BLIS-Net.

The first implementation is a pytorch implementation (code here). BLIS-Module outputs scattering features on each node, and may thus be considered as a form of message passing. This implementation may be flexibly incorporated into GNN architectures for a variety of downstream tasks. An example implentation is given in the BlisNet class from blis_legs_layer.py.

The first implementation utilizes numpy to compute the scattering moments and write them to memory (code here), after which a variety of classifiers may be trained on top of the computed scattering moments (code here).

Installation

Create a conda environment

conda create -n blis python=3.9`
conda activate blis
cd blis
pip install -e .

note: it may also be necessary to install torch-scatter

Data download (optional)

The data used in the paper may be downloaded from the following link. Please download the zip into the main project directory data directory, perhaps following something like:

rm -rf data
unzip data.zip
mv data_export data
rm data.zip

Quick Start

Pytorch implementation

A script to run the pytorch implementation is provided in scripts. From the main directory, run:

python scripts/classify_torch.py --model BlisNet --dataset synthetic --sub_dataset gaussian_pm --task_type PLUSMINUS

Numpy implementation

One script is used to first compute scattering coefficients and a second one is used to train a variety of classifiers on them. For example, one might run:

python scripts/calculate_scattering.py --scattering_type blis --wavelet_type W2 --largest_scale 4 --highest_moment 3 --dataset traffic --sub_dataset PEMS08
python scripts/classify_scattering.py --dataset=traffic --largest_scale=4 --sub_dataset=PEMS08 --scattering_type=blis --task_type=DAY --moment_list 1 --layer_list 1 2 3 --model SVC

Help

If you have any questions or require assistance using BLIS-Net, please contact us at https://krishnaswamylab.org/get-help.