Graph Machine Learning
Representation learning on graphs: message passing, pooling, and normalized adjacency first, then centrality and similarity for link prediction, then embeddings with random walks, LINE, and WL kernels. Finish by training GNN-style layers and finding communities spectrally. Take it after Graph Algorithms and Deep Learning Essentials.
Your progress
0/36 · 0%Path readiness 0%Easy 0% · Medium 0% · Hard 0%
Continue path · Adjacency List BuildWhat you will learn
- Turn graphs into tensors with adjacency, message passing, and pooling
- Measure node importance and similarity for link prediction
- Learn node embeddings with random walks, LINE, and WL kernels
- Train GNN-style layers and detect communities spectrally
Before you start
1Graph Data and Message Passing
The computational primitives: building adjacency lists and matrices, message-passing means, sum and max pooling, readout means, and GCN-style adjacency normalization.
- Adjacency List BuildEasygraph-001Graph Algorithms
- Easy
- Message Passing MeanEasygraph-181Graph Algorithms
- Graph Pooling SumEasygraph-182Graph Algorithms
- Graph Pooling MaxEasygraph-183Graph Algorithms
- Readout MeanEasygraph-184Graph Algorithms
- Easy
- Easy
Stage checkpoint
0/2 to passA checkpoint is a short retrieval test drawn from this stage: solve 2 of 3 from memory to mark the stage complete. The stage stays open either way.
2Centrality and Similarity
Score nodes and edges: degree centrality, PageRank sweeps and convergence, Adamic-Adar and Jaccard scores, cosine similarity, and preferential attachment for link prediction.
- Degree CentralityEasygraph-009Graph Algorithms
- Easy
- Medium
- Adamic-Adar ScoreEasygraph-097Graph Algorithms
- Easy
- Easy
- Easy
Stage checkpoint
0/2 to passA checkpoint is a short retrieval test drawn from this stage: solve 2 of 3 from memory to mark the stage complete. The stage stays open either way.
3Node Embeddings
Learn representations: random-walk negative sampling, Node2Vec transition probabilities, metapath counts, LINE first- and second-order losses, and Weisfeiler-Lehman hashes and kernels.
- Medium
- Medium
- Metapath CountMediumgraph-205Graph Algorithms
- Medium
- Medium
- Medium
- WL Kernel SimilarityMediumgraph-211Graph Algorithms
Stage checkpoint
0/2 to passA checkpoint is a short retrieval test drawn from this stage: solve 2 of 3 from memory to mark the stage complete. The stage stays open either way.
- LINE Second-Order LossMedium
- WL Hash Colors One RoundMedium
- WL Kernel SimilarityMedium
4GNN Training
Train and evaluate: node-classification accuracy, hits@k, attention aggregation, a full GCN layer output, graph-autoencoder and contrastive losses, GIN sum-versus-mean checks, and graph-transformer positional encodings.
- Medium
- Hits@kMediumgraph-198Graph Algorithms
- Medium
- GCN Layer OutputMediumgraph-200Graph Algorithms
- Medium
- Medium
- Easy
- Medium
Stage checkpoint
0/2 to passA checkpoint is a short retrieval test drawn from this stage: solve 2 of 3 from memory to mark the stage complete. The stage stays open either way.
- Graph Autoencoder Loss LiteMedium
- Contrastive Graph Loss LiteMedium
- Graph Transformer Positional EncodingMedium
5Spectral Methods and Communities
Global structure: Laplacian construction, spectral clustering with Fiedler signs, spectral embeddings, modularity, label propagation, and Louvain modularity gain.
- Easy
- Hard
- Hard
- Medium
- Medium
- Hard
Stage checkpoint
0/2 to passA checkpoint is a short retrieval test drawn from this stage: solve 2 of 3 from memory to mark the stage complete. The stage stays open either way.
- Modularity of a PartitionMedium
- Community Label Propagation IterationMedium
- Louvain Modularity Gain LiteHard
Capstone
CollectionClose the path with one real artifact from the catalog.
Revisit traversals, shortest paths, and connectivity before training on graphs.