Computational complexity of isometric tensor network states
The paper investigates the computational complexity of computing local expectation values in isometric tensor network states (isoTNS), finding it to be BQP-complete. It introduces strongly injective isoTNS, which possess efficient classical algorithms for expectation values when the injectivity parameter is high enough. These states exhibit interesting physical properties like long-range correlations or approximate Markov behavior depending on their injectivity level.