论文部分内容阅读
极限学习机(ELM)作为一种单隐层前馈神经网络已成为大数据分析的重要工具。与传统神经网络相比,ELM具有结构简单、学习速度快和推广性较好等优势。但是,ELM的输出权值是基于最小二乘法估计的,容易夸大离群点和噪声的影响,导致其预测性能的不稳定。提出一种新的稳健的极限学习机——基于最小一乘回归的极限学习机(LAD-ELM),而且问题被转化为线性规划,能够简单、快速求解其全局最优解。进一步将LAD-ELM应用于近红外光谱数据建模,构建了基于LAD-ELM和近红外光谱数据的乌拉尔甘草种子硬实性分析系统。与传统的方法相比,在不同光谱范围的数值实验显示了提出方法的可行性和有效性,为利用近红外光谱和ELM技术进行种子硬实性研究提供了理论依据和实用方法。
Extreme learning machine (ELM) as a single hidden layer feedforward neural network has become an important tool for big data analysis. Compared with the traditional neural network, ELM has the advantages of simple structure, fast learning speed and good generalization. However, the output weights of the ELM are estimated based on the least square method, which easily overshoots the impact of outliers and noise, resulting in unstable performance of their prediction. A new Robust Extreme Learning Machine - Least-squares Regression Based Least-Learning Machine (LAD-ELM) is proposed, and the problem is transformed into a linear programming which can solve its global optimal solution simply and quickly. The LAD-ELM was further used to model near-infrared spectral data, and a Ulaanaguan seedhardness analysis system based on LAD-ELM and near-infrared spectroscopy data was constructed. Compared with the traditional methods, numerical experiments in different spectral ranges show the feasibility and effectiveness of the proposed method, which provides a theoretical basis and practical methods for the study of hard seeds using near infrared spectroscopy and ELM techniques.