RustKernel Graph Analytics

GPU-accelerated graph analytics kernels including centrality measures, community detection, motif analysis, similarity metrics, and AML-focused analytics.

Kernels

Centrality (6 kernels)

• DegreeCentrality - Ring kernel, O(1) query
• BetweennessCentrality - Ring kernel, Brandes algorithm
• ClosenessCentrality - Ring kernel, BFS-based
• EigenvectorCentrality - Ring kernel, power iteration
• PageRank - Ring kernel, power iteration with teleport
• KatzCentrality - Ring kernel, attenuated paths

Community Detection (3 kernels)

• ModularityScore - Batch kernel
• LouvainCommunity - Batch kernel, multi-level optimization
• LabelPropagation - Batch kernel

Motif Detection (3 kernels)

• TriangleCounting - Ring kernel
• MotifDetection - Batch kernel, k-node subgraph census
• KCliqueDetection - Batch kernel

Similarity (5 kernels)

• JaccardSimilarity - Batch kernel
• CosineSimilarity - Batch kernel
• AdamicAdarIndex - Batch kernel
• CommonNeighbors - Batch kernel
• ValueSimilarity - Batch kernel (JSD/Wasserstein)

Metrics (5 kernels)

• GraphDensity - Batch kernel
• AveragePathLength - Batch kernel
• ClusteringCoefficient - Batch kernel
• ConnectedComponents - Batch kernel
• FullGraphMetrics - Batch kernel

Topology (2 kernels)

• DegreeRatio - Ring kernel, source/sink classification
• StarTopologyScore - Batch kernel, hub-and-spoke detection

Cycles (1 kernel)

• ShortCycleParticipation - Batch kernel, 2-4 hop cycle detection (AML)

Paths (1 kernel)

• ShortestPath - Batch kernel, BFS/Delta-Stepping SSSP/APSP

Graph Neural Networks (2 kernels)

• GNNInference - Message passing neural network inference
• GraphAttention - Graph Attention Network (GAT) with multi-head attention