Probabilistic tensor optimization of quantum circuits for the max-$k$-cut problem

Abstract: We propose a technique for optimizing parameterized circuits in variational quantum algorithms based on the probabilistic tensor sampling optimization. This method allows one to relax random initialization issues or heuristics for generating initial guess of variational parameters, and can be used to avoid local minima. We illustrate our approach on the example of the quantum approximate optimization algorithm (QAOA) applied to the max-$k$-cut problem based on the binary encoding efficient in the number of qubits. We discuss the advantages of our technique for searching optimal variational parameters of QAOA circuits in comparison to classical optimization methods.