A hands-on approach to modelling and analyzing constraint satisfaction problems (CSPs) by encoding them as tensor networks and contracting those networks with generic element types. Covers representing a CSP as an energy model / partition function, converting it to a tensor network, optimizing the contraction order, and reading off solution-space properties -- the partition function, solution size, number of solutions, and solution enumeration or sampling -- as well as reductions between CSPs. Taught through the Julia ecosystem (GenericTensorNetworks.jl over ProblemReductions.jl and the OMEinsum.jl contraction engine) with runnable code, worked examples such as computing an entropy constant, and analysis of the overlap gap property.
A tutorial / programming guide (physics.comp-ph) by Gao, Li and Liu that teaches, with copy-pasteable Julia code, how to model and analyze constraint satisfaction problems (CSPs) using tensor network…