极限学习机拓展研究及其在近红外光谱分析中的应用

Extreme learning machine（ELM）is a kind of single layer feedforward neural networks. Comparing with traditional neural network algorithms，it is simpler in structure，with higher learning speed and good generalization performance．Thus ELM has become a popular topic for solving big data learning, and it has been widely applied in many fields．However, for different data setting and different applications, it is used for both data classification or regression，the traditional ELM encounters some problems such as dealing with the noise data, uncertain data and high-dimension data and big data etc. Thus, in order to solve these problems and improve ELM'generalization, this project presents some new ELM models based on optimization theory and method. These extension researches for ELM include semi-supervised ELM, uncertain ELM, robust ELM, sparse ELM and ensemble ELM. Moreover, theoretical innovation and effective algorithm are used as our goal in this study. And the proposed ELM models are directly applied to the non-destructive testing of maize seeds using near-infrared (NIR) spectroscopy analysis, in order to find the quantitative relation between the NIR spectrum of maize seed and its the physiological and biochemical indexes, and interpret NIR spectra. This will promote NIR spectroscopy analysis, and on the contrary, enrich and improve ELM theory.

极限学习机是一种单隐层前馈神经网络，它较传统神经网络具有结构简单，训练速度快和良好的泛化性能等优点.因此，近年来极限学习机已成为大数据学习的热点并广泛应用。然而，但对于不同的数据集和不同的应用领域，无论极限学习机用于数据分类还是回归，传统的极限学习机还存在不足,例如在处理噪音数据、不确定性数据、高维数和大数据等问题。为了解决这些问题和进一步提高极限学习机的泛化能力，本项目以最优化理论和方法为基础，拓展现有极限学习机，构建若干个新的最优化模型及其有效算法,包括鲁棒性极限学习机、半监督极限学习机、稀疏极限学习机、不确定性极限学习机和集成极限学习机等，力求理论上的创新和算法有效性。并构建玉米种子近红外光谱无损检测系统，寻求玉米种子近红外光谱与其生理生化指标间的数量关系，进行图谱分析。这将促进近红外光谱技术发展，反之丰富和完善极限学习机和最优化理论。

传统的极限学习机在处理：噪音数据、不确定性数据、高维数和大数据等方面有待改进和完善。为了提高极限学习机的泛化能力，本项目以最优化和随机性理论为基础，以极限学习机为研究对象，以近红外光谱分析为应用背景，拓展了极限学习机理论研究和应用，构建若干个新模型及其有效算法,包括鲁棒性极限学习机、半监督极限学习机、稀疏极限学习机、光滑限学习机，不确定性极限学习机和集成极限学习机等.并证明了解的性质及建立了有效的求解算法。应用于玉米种子的近红外光谱快速无损检测，分析了玉米种子的近红外光谱与其生理生化指标间的数量关系，这将促进近红外光谱技术发展，反之丰富和完善极限学习机。