Benchmark Graph-Embedding Methods on a Climate-Network Dataset
Overview
What this challenge is about.
You receive a 200M-edge sample of the knowledge graph and a labeled entity-similarity test set (5,000 pairs with relevance labels). Benchmark three methods: a shallow embedding (Node2Vec or DeepWalk), a heterogeneous graph neural network (e.g., RGCN), and a scalable approach (GraphSAGE or a PyTorch Geometric / DGL variant). Report training cost, recall@10, and qualitative entity-neighborhood quality. Deliver the benchmark notebook, results table, and a 3-page recommendation memo.
The Brief
What you'll do, and what you'll demonstrate.
Pick the best-trade-off graph-embedding method for a 200M-edge climate knowledge graph by accuracy, cost, and qualitative neighborhood quality.
Earning criteria — what you'll demonstrate
- Apply scalable graph-embedding methods to a real heterogeneous graph
- Benchmark across accuracy and cost dimensions on a labeled test set
- Surface qualitative signal beyond aggregate metrics
- Communicate a methodology recommendation for a public-good system
Program Fit
Where this fits in your program.
Sharpens the same skills your degree expects you to demonstrate.
Skills
Skills you'll demonstrate.
Each one shows up on your verified credential.
Careers
Roles this prepares you for.
Real titles. Real skill bridges. Pick the one closest to your trajectory.
ML Researcher
Benchmarking graph-embedding methods at real scale with a written recommendation is the day-one task of an ML researcher on a knowledge-graph team.
This challenge sharpens
- graph-embeddings
- graph-neural-networks
- benchmarking
Applied AI Scientist
Connecting research methods to a public-good product surface is exactly what applied AI scientists do at mission-driven orgs.
This challenge sharpens
- graph-embeddings
- scalable-ml
- evaluation
Data Scientist
Disciplined benchmarking with qualitative inspection on a real labeled set is the bread and butter of senior data-science work.
This challenge sharpens
- benchmarking
- evaluation
- graph-embeddings