基于深度无监督学习的图像分类算法

为了取得精确的图像分类效果,一方面需要提取大量的图像特征数据进行样本分析,另一方面大量的数据又造成了维数灾难.因此,为了解决信息全面与维数灾难的矛盾,引入了深度学习.深度学习利用分层结构处理复杂的高维数据,可以完成复杂函数的逼近,是一类具有多层非线性映射的学习算法,但深度学习模型优化困难且对隐层参数敏感.针对上述问题,将无监督算法引入深度学习,这种学习方法无须人工设计特征提取数据,训练过程中样本标签是未知的.实验表明,该算法在不影响图像分类效果的前提下,可以大大降低计算复杂度,具有一定的时效性.

Research on Image Classification Algorithm Based on Deep Unsupervised Learning

In order to obtain the accurate image classification effect,on the one hand,a large amount of image feature data need to be extracted for sample analysis; on the other hand,a large amount of data cause the dimensionality disaster. Therefore,in order to solve the contradiction between comprehensive information and dimensionality disaster,deep learning is introduced. By using the hierarchical structure,deep learning can process complex high-dimensional data to complete the approximation of complex functions. It is a kind of learning algorithm with multi-layer nonlinear mapping. However,the deep learning model is difficult to optimize and sensitive to hidden layer parameters. In view of the above problems,the idea of unsupervised algorithm is introduced into deep learning. This learning method does not need to design features to extract data,so the sample labels are unknown during training. Experiments show that this algorithm can reduce the computational complexity greatly and has certain timeliness under the premise of not affecting the image classification effect.