Trends in Classical Angle Optimization of ma-QAOA

Quantum approximation optimization algorithm (QAOA) is a variational quantum algorithm that has been well studied due to its application of solving combinatorial optimization problems. One such problem of interest is the MaxCut problem, where given a graph $G$, what is the best way to partition the vertices of $G$ such that the number of edges of $G$ that have a vertex in each partition is maximized. However, due to the large qubit and circuit depth requirements, implementing QAOA is not practical on NISQ devices. Multi-angle QAOA (ma-QAOA) is a variant of QAOA in which all clauses receive angles. It has been observed that 1-ma-QAOA outperforms 3-QAOA on small graphs. We explore whether $p$-ma-QAOA performs better than $(p+2)$-QAOA for larger $p$. We will also discuss how ma-QAOA circuits are, in general, shallower than QAOA circuits and how choice of initial classical optimizer seed for angle optimization affects algorithm output.