半导体光电, 2019, 40 (1): 112, 网络出版: 2019-03-25  

改进鲸鱼算法优化核极限学习机在水质光谱分析中的应用

Application of Kernel Extreme Learning Machine Optimized by Improved Whale Algorithm in Water Quality Spectrum Analysis
作者单位
中国计量大学 机电工程学院, 杭州 310018
摘要
为提高水质光谱分析模型的学习速度与预测精度, 采用核极限学习机对水质光谱进行建模, 并提出一种具有动态惯性权重的改进鲸鱼优化算法对模型进行参数优化。由于极限学习机的输入权值矩阵和偏置是随机生成的, 故引入核方法以减小其输出权值矩阵的波动; 将鲸鱼优化算法中的惯性权重在非线性递减的基础上引入随机因子, 通过动态调整惯性权重以平衡算法的全局搜索能力与局部开发能力。与传统优化模型进行了对比实验, 实验结果表明: 基于该方法所建模型具有更高的预测精度, 而在相同的学习迭代次数下, 核极限学习机的运行时间相对于传统算法约下降50%, 且改进鲸鱼优化算法能够以更快的收敛速度使模型达到全局最优。
Abstract
In order to improve the learning speed and prediction accuracy of water quality spectrum analysis model, kernel extreme learning machine was used to model the water quality spectrum, and an improved whale optimization algorithm with dynamic inertia weight was proposed to optimize the model parameters. Since the input weight matrix and bias of the extreme learning machine are generated randomly, the kernel method was introduced to reduce the fluctuation of the output weight matrix. The inertia weight of whale optimization algorithm was introduced into the random factor based on nonlinear decreasing, and the dynamic adjustment of inertia weight was used to balance the global search capability and the local development ability. Compared with the traditional optimization model, the experimental results show that the model based on this method has higher prediction accuracy, and under the same number of learning iterations, the running time of the kernel extreme learning machine is about 50% lower than traditional algorithm, and the improved whale optimization algorithm enables the model to achieve global optimum at a faster convergence rate.

林春伟, 郭永洪, 何金龙. 改进鲸鱼算法优化核极限学习机在水质光谱分析中的应用[J]. 半导体光电, 2019, 40(1): 112. LIN Chunwei, GUO Yonghong, HE Jinlong. Application of Kernel Extreme Learning Machine Optimized by Improved Whale Algorithm in Water Quality Spectrum Analysis[J]. Semiconductor Optoelectronics, 2019, 40(1): 112.

关于本站 Cookie 的使用提示

中国光学期刊网使用基于 cookie 的技术来更好地为您提供各项服务,点击此处了解我们的隐私策略。 如您需继续使用本网站,请您授权我们使用本地 cookie 来保存部分信息。
全站搜索
您最值得信赖的光电行业旗舰网络服务平台!