基于高斯过程隐变量模型的图像数据降维算法

A Novel Dimension Reduction Algorithm of Image Data Based on Gaussian Process Latent Variable Mode

针对传统谱算法在数据降维计算复杂度高的缺点,提出一种基于高斯过程隐变量模型的图像数据降维算法。首先,通过高斯过程(Gaussian Process,GP)建立图像数据的概率模型,得到图像数据的隐变量模型;其次,利用概率最大化原则得到最优超参数,通过最优超参数求取最优数据降维结果;最后,实现图像数据降维。选取Yale,ORL两类数据集与传统算法进行人脸识别对比实验,实验结果表明:所提出的算法针对图像数据降维问题有较好的效果,结合支持向量机算法,可有效地对人脸图像进行识别,且有较高的识别率,从而体现出算法对高维数据降维的准确性。

For the characteristics of image data dimension reduction, a novel dimension reduction algrithm based on Gaussian Process Latent Variable Mode （GP-LVM） is proposed. Firstly, the probabilistie model of image data is established by the Gaussian Process （GP） , and the LVM of the data can be gotten. Secondly, the super-parameter can be gotten by the principle of maximum probability, and the optimization results of data dimension reduction can be gotten by the optimization super-parameter. Finally, the image data di- mension reduction can be achieved. The two classes of data sets are selected as the experimental data, which consist of Yale and ORL. The experiment results show that the proposed method has a great effect to reduce dimension of image data. The proposed algrithm com- bine GP-LVM with the support vector machine, and the face recognition can be achieved. It has a great recognition rate, and the pro- posed algrithm has a accuracy for the high dimension data.