Hyperbolic Convolution via Kernel Point Aggregation

This repo provides the official implementation of the HKConv from the following paper.

Hyperbolic Convolution via Kernel Point Aggregation
Eric Qu, Dongmian Zou
arXiv: https://arxiv.org/abs/2306.08862

Environment

The code is tested on Python 3.9, PyTorch 1.12.1, and CUDA 11.3.

First, install PyTorch from https://pytorch.org/. Then, install the other dependencies by

pip install -r requirements.txt

Training

The training scripts are in scripts\. You can use them by

bash scripts/experiment/dataset.experiment.sh

File Descriptions

data/: Datasets
distribution/: Hyperbolic Distributions
kernels/: Kernel generation
layers/: Hyperbolic layers
manifolds/: Manifold calculations
models/: GNN models
optim/: Optimization on manifolds
scripts/: Training scripts
utils/: Utility files
train.py: Training scripts

Our HKConv is located in layers/hyp_layers.py.