An Adaptive Neural Network Algorithm with Quasi Opposition-Based Learning for Numerical Optimization Problems

The structure of artificial neural networks and the biological nervous systems serve as the foundation for the creation of the neural network algorithm (NNA). The robust global search capability of NNAs makes it an effective tool for solving a wide range of complex optimization problems. Unfortunately, its limited relevance to many optimization problems is due to its poor exploitation, weak convergence, and tendency to fall into local optima. The paper's goal is to introduce an enhanced version of the NNA known as the adaptive quasi-opposition-based neural network algorithm (AQOBNNA) in order to overcome these issues. The quasi-opposition-based learning (QOBL) and an adaptive strategy are combined in this suggested algorithm, where the adaptive technique is added to determine whether or not to use QOBL. The QOBL technique replaces a random search individual with the best one throughout the position update phase in order to enhance exploitation and increase exploration capabilities. The performance of the suggested AQOBNNA is assessed using a set of 23 traditional benchmark functions and compared with a number of current methods. It is evident from the experimental data that AQOBNNA performs better overall and outperforms all the algorithms that were examined.