Abstract
Quantum Approximate Optimization Algorithm (QAOA) is one of the mostpromising candidates to achieve the quantum advantage in solving combinatorialoptimization problems. The process of finding a good set of variationalparameters in the QAOA circuit has proven to be challenging due to multiplefactors, such as barren plateaus. As a result, there is growing interest inexploiting parameter transferability, where parameter sets optimized for oneproblem instance are transferred to another that could be more complex eitherto estimate the solution or to serve as a warm start for further optimization.But can we transfer parameters from one class of problems to another?Leveraging parameter sets learned from a well-studied class of problems couldhelp navigate the less studied one, reducing optimization overhead andmitigating performance pitfalls. In this paper, we study whether pretrainedQAOA parameters of MaxCut can be used as is or to warm start the MaximumIndependent Set (MIS) circuits. Specifically, we design machine learning modelsto find good donor candidates optimized on MaxCut and apply their parameters toMIS acceptors. Our experimental results show that such parameter transfer cansignificantly reduce the number of optimization iterations required whileachieving comparable approximation ratios.