gnn-algorithm

(WWW 2021) Source code of PC-GNN

Meta graph convolutional neural network-assisted resilient swarm communications

A Multi-relational Graph-Based Twitter Account Detection Benchmark

This repo contains the code and data of "Graph Matching with Bi-level Noisy Correspondence".

Solving pursuit-evasion problems on graphs using Reinfocement Learning and GNNs

Pytorch Geometric implementation of the "Gravity-Inspired Graph Autoencoders for Directed Link Prediction" paper.

Synthetic multiplex terrorist networks and multi-task GNNs for high-value target detection and link prediction experiments.

A Python library for building AI agents that leverage the full power of Google Antigravity.

Public demonstration interface and vulnerability scanner for GNN-based smart contract auditing, featuring static risk detection and Solana Anchor safety registry integration.

Data and code of CADGL: Context-Aware Deep Graph Learning for Predicting Drug-Drug Interactions

SelfGNNplus achieves up to 5.97% improvement in Hit Rate and 6.41% in NDCG compared to the best baseline models. Ablation studies highlight the critical role of interval-level dependencies, revealing their substantial impact on recommendation accuracy. The study also examines the effects of key hyperparameters on model performance and computation.

"Structural Hazards in Global Production: A Graph Neural Network Approach to Sectoral Survival"** by Diego Vallarino (2025), as a Working Paper status

ARC-AGI Hybrid Architecture A standalone neural-symbolic solver. Features : Perception: Multi-modal Transformers for grid-to-latent mapping. Reasoning: GNNs for entity-relational graph processing. Synthesis: Efficient DSL search guided by neural embeddings and D4 group symmetry.

Repository for FAME and A-FAME used to perform bias mitigation when using GNNs

This project demonstrates parallelization techniques for Graph Neural Networks (GNNs) using: CUDA for GPU acceleration MPI (mpi4py) for distributed computing Python Multiprocessing for parallel processing The implementation uses the PubMed dataset from PyTorch Geometric and a 2-layer GCN model.

Graph Neural Networks with Heterogeneous Meta-Path Learning for Cold-Start User Recommendation