Ensembles of fuzzy cognitive map classifiers based on quantum computation

Abstract

Fuzzy cognitive maps (FCMs) have been widely employed in the dynamic simulation and analysis in the complex systems. While a novel classifier model based on FCMs (FCMCM) was proposed in our former work, the obvious bottleneck of the genetic leaning algorithm used in FCMCM is its irksome efficiency, in particular, low speed in cross over and mutation delay in global convergence. Moreover the lack of the necessary robustness of a single FCMCM limits its generalization. To this end, a quantum computation based ensemble method FCMCM_QC is proposed to address the scalability problem, which employs a novel evolutionary algorithm inspired by quantum computation. The FCMCM_QC effectively uses the concept and principle of quantum computation to facilitate the computational complexity of genetic optimization for the FCMCM and reasonably selects classifiers with better performance for efficient ensembles. The experimental studies demonstrate the quality of the proposed FCMCM_QC in generally used UCI datasets, and the simulation results prove that the FCMCM_QC does enhance the speed of the convergence with high efficiency and good quality.