Quantum-inspired optimization algorithm with adaptive correction of energy position: Methodology and a case study

Abstract

Efficient and stable global optimizers constitute a noteworthy arena of academic study and real-world applications. Since Multi-scale Quantum Harmonic Oscillator Algorithm inspired by the quantum motion for solving optimization problems was proposed, considerable contributions regarding this algorithm have been achieved in recent years. Nevertheless, issues such as the aggregation effect during sampling as well as recurrence and blindness in random searches hinder the performance of the algorithm. Motivated by this situation, a variant of Multi-scale Quantum Harmonic Oscillator Algorithm is put forward to improve the efficiency of the system convergence while maintaining the solution diversity. The measurement of the solution position through the collapse of the quantum state to the classical state is realized by means of quantum Monte Carlo simulations, and the energy position is established as a metric for energy observation. Then, the adaptive correction of the energy position is explored to improve algorithm performance. The core idea of our mechanism is to adaptively guide the candidate solutions toward convergence to the ground state by means of attractive factors based on the relationship among the energy positions of several reference points. Experimental results obtained on the CEC2013 benchmark functions and a real-world application indicate that the performance of our scheme is competitive and that it achieves prominence among the compared algorithms as the dimensionality increases.