Overcoming Performance Inequity in Quantum Dynamics Framework for Optimization of Symmetric and Asymmetric Double-Well Function

In recent years, a study focused on the quantum dynamics framework (QDF) process has revealed significant, random structural bias in QDF’s optimization of symmetric and asymmetric forms of double-well functions (DWF). Such bias means that the algorithm tends to favor specific solutions in the search space while ignoring other possible solutions. This tendency may cause the algorithm to become trapped in local optima, preventing it from finding global optima and ultimately compromising its performance. To address this performance gap, this paper proposes a dimension-decoupling sampling mechanism and random-neighborhood sampling mechanism. The proposed mechanism effectively reduces the coupling between dimensions during the optimization of the objective function and increases the search space diversity, thereby mitigating the structural bias of the algorithm. It successfully solves the issue of performance degradation observed in the optimization of symmetric double-well functions. We refer to the modified algorithm as the dimension-decoupling and random-neighborhood quantum dynamics framework (DR-QDF).