AI Glossary
The complete dictionary of Artificial Intelligence
Louvain Algorithm
Heuristic method for community detection based on iterative modularité optimization, using a multi-level approach to identify hierarchical community structures in large graphs.
Girvan-Newman Algorithm
Hierarchical community detection method that progressively identifies communities by removing edges with the highest betweenness centrality, thus revealing the intrinsic structure of the graph.
Overlapping Community Detection
Partitioning approach allowing nodes to belong simultaneously to multiple communities, better adapted to social networks where entities can have multiple affiliations.
Spectral Clustering Method
Algebraic technique using the eigenvalues and eigenvectors of the graph's Laplacian to project nodes into a low-dimensional space where clustering becomes more discriminative.
Leiden Algorithm
Improvement of the Louvain algorithm guaranteeing well-connected communities and offering faster convergence while preserving the quality of community detection.
Weighted Graph
Data structure where each edge has a numerical value representing the intensity or frequency of interactions between nodes, directly influencing community detection algorithms.
Conductance Metric
Quality indicator of a community measuring the ratio between the number of outgoing edges and the total volume of the community, with low values indicating well-isolated communities.
Hierarchical Graph Clustering
Detection approach building a hierarchy of communities through successive merging (agglomerative) or progressive division (divisive) to reveal multi-scale structures in the network.
Infomap
Algorithm based on information theory that identifies communities by minimizing the description length of random walks on the graph, interpreting communities as modules that compress information.
Walktrap
Detection method exploiting similarities between short random walks starting from each node, grouping nodes with similar walking behaviors into the same communities.
Dynamic community detection
Set of techniques designed to identify and track the evolution of communities in temporal graphs, capturing phenomena of group formation, dissolution, and merging over time.
K-means algorithm on graphs
Adaptation of K-means clustering to graph structures using distance metrics based on shortest paths or node embeddings to partition the network.
Label propagation method
Semi-supervised algorithm where each node adopts the majority label of its neighbors, converging rapidly to communities without requiring complex parameters or costly iterative optimization.
Kernighan-Lin algorithm
Binary partitioning heuristic that optimizes the number of inter-community edges by successive exchanges of node pairs between the two partitions to minimize the cut.
Community dendrogram
Tree representation of hierarchical relationships between communities, visualizing how groups merge or divide at different resolution levels of the partitioning.
Multi-scale community detection
Analysis paradigm that simultaneously identifies community structures at different resolutions, recognizing that social networks often present a complex hierarchical organization.
BigCLAM Algorithm
Method for detecting overlapping communities based on a generative model where each community is represented by a vector in a latent space, optimizing the likelihood of node affiliations.