iMoonLab/DeepHypergraph

A pytorch library for graph and hypergraph computation.

What it solves

DHG (DeepHypergraph) provides a unified deep learning framework for working with both standard graphs (low-order structures) and hypergraphs (high-order structures). It simplifies the process of designing and implementing Graph Neural Networks (GNNs) and Hypergraph Neural Networks (HGNNs) by providing standardized tools for message passing, structure manipulation, and model evaluation.

How it works

Built on PyTorch, DHG integrates spectral-based operations (such as Laplacian-based smoothing) and spatial-based operations (message passing) directly into the structure objects. It supports a wide range of message-passing patterns, including vertex-to-vertex, vertex-to-hyperedge, and hyperedge-to-vertex. The library also includes an Auto-ML module powered by Optuna to automatically tune model hyperparameters and structure configurations.

Who it’s for

It is designed for researchers and developers working on geometric deep learning, specifically those implementing GNNs or HGNNs for tasks like vertex classification and user-item recommendation.

Highlights

  • High-Order Message Passing: Supports beyond-pair-wise communication on hypergraph structures.
  • PyTorch Integration: Fully compatible with the PyTorch ecosystem, allowing any PyTorch-based model to be integrated.
  • Structure Conversion: Includes functions to convert between standard graphs and hypergraphs to exploit high-order connections.
  • Built-in Visualization: Tools to visualize graphs, hypergraphs, directed graphs, and bipartite graphs.
  • Auto-ML Capabilities: Integrated Optuna support for automatic hyperparameter and structure tuning.
  • Comprehensive Resources: Includes a large library of public datasets, SOTA models, and evaluation metrics.

Related

  • Project
    TorchSpatiotemporal/tsl

    A PyTorch-based library for neural spatiotemporal forecasting that provides a unified framework for data preprocessing, model prototyping, and Graph Neural Network implementation.

  • Project
    pyg-team/pytorch_geometric

    PyTorch Geometric (PyG) is a PyTorch‑based library for building, training, and scaling graph neural networks. It offers a unified API, dozens of ready‑made GNN layers, data loaders for large and heterogeneous graphs, and easy extensibility via a `MessagePassing` base class.

  • Project
    pytorch/pytorch

    A Python-based machine learning framework that provides GPU-accelerated tensor computation and a dynamic autograd system for building deep neural networks.

  • Project
    pytorch-tabular/pytorch_tabular

    A high-level PyTorch framework that provides a unified interface for applying various deep learning architectures to tabular data.

  • Project
    Lightning-AI/pytorch-lightning

    A deep learning framework that abstracts away the engineering boilerplate of PyTorch, enabling researchers to pretrain and finetune AI models at scale across various hardware.