Zhang Yinchao, Li Ting, Chen He, Chen Siying, Guo Pan, Li Yi
Appl Opt. 2019 Mar 20;58(9):2340-2349. doi: 10.1364/AO.58.002340.
An optimized dimensionality reduction technique is proposed as the improved continuous locality preserving projection (ICLPP), which was developed by modifying and optimizing the weighting functions and weighting factors of the continuous locality preserving projection (CLPP) algorithm. With only one adjustable parameter, this optimized technique not only enhances CLPP's capability of maintaining the continuity of the massive data, but also results in better simplicity and adaptability of the algorithm. In this paper, the performance of ICLPP is validated through quantification analysis of the adulteration of extra virgin olive oil (EVOO) with low-cost oils based on laser-induced fluorescence spectroscopy. Through cross validation and comparative studies, ICLPP, combined with the regression algorithm, is employed to predict and screen adulteration in EVOO, and is found to generally outperform other state-of-the-art dimensionality reduction algorithms, especially for prediction of adulterants at low level (<10%). It is evidenced that the ICLPP-based framework is superior in detecting adulteration by using spectral data.
提出了一种优化的降维技术，即改进的连续局部保持投影（ICLPP），它是通过修改和优化连续局部保持投影（CLPP）算法的加权函数和加权因子而开发的。该优化技术仅具有一个可调参数，不仅增强了CLPP保持海量数据连续性的能力，还使算法具有更好的简洁性和适应性。本文通过基于激光诱导荧光光谱对特级初榨橄榄油（EVOO）与低成本油掺假的定量分析，验证了ICLPP的性能。通过交叉验证和对比研究，将ICLPP与回归算法相结合，用于预测和筛选EVOO中的掺假情况，发现其总体性能优于其他现有先进的降维算法，特别是对于低水平（<10%）掺假物的预测。结果表明，基于ICLPP的框架在利用光谱数据检测掺假方面具有优势。
