opencog/matrix

Provides statistical analysis tools (correlation, mutual information, similarity metrics) for discovering patterns in large sparse graphs by treating similar subgraphs as vectors, explicitly optimized for AtomSpace's hypergraph database to handle million-dimensional vectors that would exhaust dense vector libraries like SciPy. Built from scratch rather than wrapping existing tools, it enables matrix operations directly on graph structures without materializing sparse data, forming the foundation for the Learn project's natural language analysis and supporting symbolic AI reasoning through graph adjacency matrix representations.