ATTENTION HORAIRE INHABITUEL
Stefanos Kourtis (Boston University)
Solving constrained counting problems with tensor networks
In this talk, I will present newly developed physics-inspired methods for the solution of counting constraint satisfaction problems (#CSPs). #CSP instances can be reformulated as interacting models whose zero-temperature partition function represents the volume of the solution manifold. I will introduce practical methods to compute such partition functions based on tensor network contraction. In this formulation, computational complexity can be viewed as a manifestation of quantum entanglement, and controlling the growth of entanglement throughout tensor network contraction can yield a significant computation speedup. Using some hard counting problems as benchmarks, I will demonstrate that tensor network methods can be a useful tool for solving some hard classes of #CSPs. I will conclude with an outline of ongoing work on extensions of this framework, such as the simulation of existing and near-term quantum circuits.