改进鲸鱼算法及其在浅层神经网络搜索中的权值阈值优化

Improved whale optimization algorithm and its weights and thresholds optimization in shallow neural architecture search

为设计出简便高效的方法搜索最优神经网络结构,提出一种改进鲸鱼优化算法的浅层神经网络搜索方法.该方法首先通过模拟鲸鱼狩猎的个体偏好行为和鲸鱼群位置移动的非线性权值更新机制对传统鲸鱼优化算法进行改进;然后将改进鲸鱼优化算法作为浅层BP神经网络结构搜索策略,构建基于浅层BP神经网络的最优网络结构的权值阈值搜索优化方法.数值实验结果表明,改进的鲸鱼优化算法不仅在求解不同维复杂函数上具有良好的寻优性能,而且通过改进鲸鱼优化算法搜索得到的最优浅层BP神经网络结构在回归任务中具有更好的预测精度和泛化性能.

In order to design a simple and efficient method to search for the optimal neural network architecture,we propose an improved whale optimization algorithm and its shallow neural architecture search’s weights and thresholds optimization.This method first improves the traditional whale optimization algorithm by simulating the individual preference behavior of whale hunting and the nonlinear weight update mechanism of whale group position movement;Then,the improved whale optimization algorithm is used as a shallow BP neural network architecture search strategy,and a architecture based on the optimization method of weight threshold search for the optimal network architecture of the shallow BP neural network is constructed.The numerical experimental results show that the improved whale optimization algorithm not only has good optimization performance in solving complex functions with different dimensions,but also the optimal shallow BP neural network architecture searched by the improved whale optimization algorithm has better prediction accuracy and generalization performance in regression tasks.