Applying Entropy Quantum Computing to Non-Convex Optimization Problems

This talk provides an overview of QCi’s photonic computing engine which leverages nonlinear optical systems in order to perform computation. The system is built for non-convex optimization. Our solver uses weakly-coherent single photon states to drive a non-Hermitian / open quantum system. We show that our new paradigm of computing can efficiently solve NP-Hard problems. We further demonstrate the utility of this paradigm by showing that in addition to optimizing binary (Ising) objective functions with qubit variables, our machine is capable of optimizing integer (Potts) objective functions using qudit variables. Finally we demonstrate the utility of this optical hardware in effectively modeling two-body interactions. Our computing paradigm can naturally optimize configurations involving up to five-body interaction terms. We conclude by showing applications of our hardware to problems such as: graph cut and graph clique problems, ensemble learning and boosting, fixed-backbone protein design, and non-negative matrix factorization.