A Method to Compute QAOA Fixed Angles

QAOA (Quantum Approximate Optimization Algorithms) is one of the most promising algorithms of Noisy Intermediate Scale Quantum (NISQ) era. The standard approach to QAOA involves the use of a hybrid quantum-classical optimization, although this approach was not considered as the main one in the original paper on QAOA. Recently, a new approach has emerged based on the hypothesis that optimal circuit parameters (angles) are close for a wide class of problems. However, the search for fixed angles itself remains a challenge with different approaches. We propose one specific method based on the use of a fixed training set and the special metric associated with increasing the probability of a correct answer. We carry out the analysis of the proposed method performance on the unweighted Max-Cut problems and random weighted QUBO (Quadratic Unconstrained Binary Optimization) problems of the special type.