palash1992/GEM

Health v2 · maintenance only

Adoption not part of the score

1288 stars · 357 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

GEM is a Python library offering a unified framework for static graph embedding algorithms, including node2vec, HOPE, SDNE, Graph Factorization, Laplacian Eigenmaps, and Locally Linear Embedding. It ships evaluation utilities for link prediction, graph reconstruction, node classification, and visualization, with a C++/Boost backend for faster execution.

Use cases

• learn low-dimensional node embeddings for a network
• compare graph embedding algorithms under a common framework
• predict missing links in a graph
• visualize a network in 2D via its embeddings
• classify nodes using embedding features
• reconstruct graph edges from learned embeddings

When to choose

• You want to run and benchmark multiple classic static graph embedding methods through one consistent Python API
• You need built-in evaluation tasks such as link prediction, node classification, and graph reconstruction to judge embedding quality
• You are doing research or a survey on graph/network embedding techniques and want an extensible framework for adding new methods

When to avoid

• You need dynamic or evolving graph embeddings (use the companion DynamicGEM library instead)
• You need a large-scale, GPU-accelerated or production-grade graph learning stack (PyTorch Geometric, DGL)
• You require knowledge-graph embeddings with entity/relation features (TransE-style) or attributed graph methods
• You need a library with frequent recent updates and long-term maintenance guarantees

Facets

library · maturity maintenance

machine-learning benchmarking data-visualization machine-learning data-science python cross-platform graph-embedding network-embedding node2vec link-prediction node-classification graph-reconstruction graph-representation-learning static-graphs research-library algorithms

Member repositories

For agents

markdown · JSON · MCP: product_card(name="palash1992/GEM")