UC San Diego

Biological, and complex systems more broadly, are characterized by their many interacting degrees of freedom. Despite their apparent complexity and scale, the microscopic constituents of these systems often act in concert to produce emergent macroscopic behavior with vastly reduced dimension. Through the lens of geometry and statistical physics, we present a collection of novel approaches to dimensionality reduction in complex systems identify their correct low dimensional description. We take a threefold approach based on embedding methods, matrix factorization, and neural compression.