Module graph_kernels 

Graph Kernel Approximations

This module implements various graph kernel approximation methods for analyzing graph-structured data such as molecular graphs, social networks, and other relational data structures.

Key Features

• Random Walk Kernels: Count common random walks between graphs
• Shortest Path Kernels: Compare shortest path distributions
• Weisfeiler-Lehman Kernels: Graph isomorphism-based kernels
• Subgraph Kernels: Count common subgraph patterns
• Graphlet Kernels: Count small connected subgraphs
• Graph Laplacian Kernels: Use graph Laplacian spectrum

Mathematical Background

Graph kernel between graphs G₁ and G₂: K(G₁, G₂) = Σ φ(G₁)[f] * φ(G₂)[f]

Where φ(G)[f] is the feature map counting occurrences of feature f in graph G.

References

• Vishwanathan, S. V. N., et al. (2010). Graph kernels
• Weisfeiler, B., & Lehman, A. A. (1968). The reduction of a graph to canonical form

Structs

FittedRandomWalkKernel

Fitted random walk kernel FittedRandomWalkKernel

FittedShortestPathKernel

Fitted shortest path kernel FittedShortestPathKernel

FittedSubgraphKernel

Fitted subgraph kernel FittedSubgraphKernel

FittedWeisfeilerLehmanKernel

Fitted Weisfeiler-Lehman kernel FittedWeisfeilerLehmanKernel

Graph

Simple graph representation Graph

RandomWalkKernel

Random walk kernel for graphs RandomWalkKernel

ShortestPathKernel

Shortest path kernel for graphs ShortestPathKernel

SubgraphKernel

Subgraph kernel that counts common subgraph patterns SubgraphKernel

WeisfeilerLehmanKernel

Weisfeiler-Lehman kernel for graphs WeisfeilerLehmanKernel