joisino/gnnrecover
Code for "Graph Neural Networks can Recover the Hidden Features Solely from the Graph Structure" (ICML 2023)
This project explores how Graph Neural Networks (GNNs) can create meaningful data about individual items in a network, even when you start with no specific information about those items. It takes only the connections between items (the graph structure) and outputs useful, newly generated characteristics for each item. This is valuable for data scientists, machine learning engineers, and researchers working with complex networked data where explicit item features are missing or uninformative.
No commits in the last 6 months.
Use this if you need to extract meaningful information from network structures when your individual data points lack clear descriptive features.
Not ideal if your existing node features are already rich and highly informative, as this project focuses on scenarios with uninformative or absent features.
Stars
24
Forks
2
Language
Python
License
MIT
Category
Last pushed
Jun 01, 2023
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/embeddings/joisino/gnnrecover"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
eliorc/node2vec
Implementation of the node2vec algorithm.
mims-harvard/decagon
Graph convolutional neural network for multirelational link prediction
mims-harvard/nimfa
Nimfa: Nonnegative matrix factorization in Python
ferencberes/online-node2vec
Node Embeddings in Dynamic Graphs
claws-lab/jodie
A PyTorch implementation of ACM SIGKDD 2019 paper "Predicting Dynamic Embedding Trajectory in...