GConvNet 0.0.3

GConvNet

GConvNet

GConvNet is a PyTorch module for graph convolution using triplet-based edge structures and dynamic node filtering. It enables higher-order neighborhood modeling via (3×1) convolution over node triplets, and selectively retains important nodes based on degree centrality.

Features

• Triplet generation from sequential edges
• (3×1) convolution on node feature triplets
• Node filtering by degree
• Simple integration with PyTorch models

Installation

Install with pip:

pip install GConvNet

Usage

import torch
from GConvNet import GConvNet

x = torch.rand(8, 4)
edge_index = torch.tensor([[0, 1, 2, 3], [1, 2, 3, 4]])
model = GConvNet(4, 8)
x_out, edge_index_out = model(x, edge_index)