Search

The success of the supervised learning process for feedforward neural networks, especially multilayer perceptron neural network (MLP), depends on the suitable configuration of its controlling parameters (i.e., weights and biases). Normally, the gradient descent method is used to find the optimal values of weights and biases. The gradient descent method suffers from the local optimal trap and slow convergence. Therefore, stochastic approximation methods such as metaheuristics are invited. Coronavirus herd immunity optimizer (CHIO) is a recent metaheuristic human-based algorithm stemmed from the herd immunity mechanism as a way to treat the spread of the coronavirus pandemic. In this paper, an external archive strategy is proposed and applied to direct the population closer to more promising search regions. The external archive is implemented during the algorithm evolution, and it saves the best solutions to be used later. This enhanced version of CHIO is called ACHIO. The algorithm is utilized in the training process of MLP to find its optimal controlling parameters thus empowering their classification accuracy. The proposed approach is evaluated using 15 classification datasets with classes ranging between 2 to 10. The performance of ACHIO is compared against six well-known swarm intelligence algorithms and the original CHIO in terms of classification accuracy. Interestingly, ACHIO is able to produce accurate results that excel other comparative methods in ten out of the fifteen classification datasets and very competitive results for others. [ABSTRACT FROM AUTHOR]

The selection process is the most attractive operator in the optimization algorithms. It normally mimics the natural selection of survival of the fittest principle. When the selection is too greedy, the selection pressure will be high and therefore the search becomes biased toward exploitation. In contrast, when the selection has a tendency to be random, the selection pressure will be low and thus the exploration is observed more. The selection process in Grey Wolf Optimizer (GWO) tends to be too greedy since the search is driven by the three best solutions. In this paper, different selection methods extracted from other evolutionary algorithms (EAs) are investigated for GWO. Along with the original selection method of GWO called greedy-based selection method, five others selection methods which are proportional, tournament, universal sampling, linear rank, and random selection methods are investigated. Accordingly, six versions of GWO are proposed which are Greedy-based GWO (GGWO), Proportional-based GWO (PGWO), Tournament-based GWO (TGWO), Universal sampling-based GWO (UGWO), Linear rank-based GWO (LGWO), Random-based GWO (RGWO). The six versions are evaluated using 23 test functions circulated in the literature with different characteristics and complexity. The sensitivity analysis of the control parameters of some new versions are discussed. Interestingly, TGWO achieved the best results for almost all benchmark functions. This proves that the selection methods have a high impact on the performance of GWO. [ABSTRACT FROM AUTHOR]

This paper presents a novel meta-heuristic algorithm so-called White Shark Optimizer (WSO) to solve optimization problems over a continuous search space. The core ideas and underpinnings of WSO are inspired by the behaviors of great white sharks, including their exceptional senses of hearing and smell while navigating and foraging. These aspects of behavior are mathematically modeled to accommodate a sufficiently adequate balance between exploration and exploitation of WSO and to assist search agents to explore and exploit each potential area of the search space in order to achieve optimization. The search agents of WSO randomly update their position in connection with best-so-far solutions, to eventually arrive at the optimal outcome. The performance of WSO was comprehensively benchmarked on a set of 29 test functions from the CEC-2017 test suite for several dimensions. WSO was further applied to solve the benchmark problems of the CEC-2011 evolutionary algorithm competition to prove its reliability and applicability to real-world problems. A thorough analysis of computational and convergence results was presented to shed light on the efficacy and stability levels of WSO. The performance score of WSO in terms of several statistical methods was compared with 9 well-established meta-heuristics based on the solutions generated. Friedman's and Holm's tests of the results showed that WSO revealed reasonable solutions, in terms of global optimality, avoidance of local minima and solution quality, compared to other existing meta-heuristics. [ABSTRACT FROM AUTHOR]