Binary QUATRE Using Time-varying Transfer Functions

Abstract

QUasi-Affine TRansformation Evolutionary (QUATRE) is an excellent algorithm that generalized differential evolution algorithm to matrix form. There are two approaches of Binary QUATRE (BQUATRE) algorithm that convert the continuous search space to the binary one. The convergence process of meta-heuristic algorithms can be regarded as two stages: exploration and exploitation. Generally speaking, transfer function plays an important role in the binary metaheuristic algorithms. Therefore, four families of transfer functions are converted into the time-varying form to balance the exploration and exploitation. In particular, a new timevarying formula for the V-shaped, U-shaped and Z-shaped transfer functions is designed in this manuscript. Then, two new approaches of BQUATRE algorithm using the timevarying transfer functions are proposed. In addition, a wrapper feature selection algorithm is designed based on the proposed time-varying BQUATRE. In order to verify the performance of the proposed algorithms, the UCI repository is used in the experiment and the results show that the proposed algorithms are superior to the original BQUATRE and the state-of-the-art algorithms.