面向图像数据集的高斯过程分类

Gaussian process classification for image datasets

高斯过程分类是近年机器学习领域引起广泛关注的一类有监督的学习算法。该算法在高斯过程的先验假设下,以后验概率最大化的为目标,获得对新样本的预测值及属于该值的概率。针对图像数据的特性,提出一种将高斯过程应用于图像分类的方法,同时在此基础上给出对图片进行排序的一种方案。在公开的图像数据集上进行了实验,并与支持向量机分类器进行对比,证实了其有效性,为改进图像分类技术提供一条可供参考的途径。

Gaussian process classification receives increased attention in the machine learning community over the past decade.It maximizes the posterior probability based on the Gaussian process prior assumption and obtains predictive probability on unlabeled samples.Implemented Gaussian process method for image classification is proposed in this paper and a strategy is given for image ranking.The algorithm is tested on several well-known object category datasets.Compared with those produced by support vector machines,it verifies the effectiveness of the proposed method and provides a new approach to improve the image classification.