基于非线性降维多项式逻辑斯蒂回归的图像/非图像数据的分类与识别

在面向大规模复杂数据的模式分类和识别问题中,绝大多数的分类器都遇到了维数灾难这一棘手的问题.在进行高维数据分类之前,基于监督流形学习的非线性降维方法可提供一种有效的解决方法.利用多项式逻辑斯蒂回归方法进行分类预测,并结合基于非线性降维的非监督流形学习方法解决图像以及非图像数据的分类问题,因而形成了一种新的分类识别方法.大量的实验测试和比较分析验证了本文所提方法的优越性.

Classification and recognition of image/non-image data based on multinomial logistic regression with nonlinear dimensionality reduction

In pattern classification and recognition oriented to massively complex data most classifiers suffer from the curse of dimensionality. Manifold learning based nonlinear dimensionality reduction (NLDR) methods provide a good preprocessing to reduce dimensionality before applying any classification method on high dimensional data. Multinomial logistic regression (MLR) can be used to predict the class membership of feature data. In this study several unsupervised NLDR methods are employed to reduce dimensions of the data and the MLR is used for class prediction of image/non-image data so that a new method of classification and recognition oriented to massively complex image/non-image data is proposed based on multinomial Logistic regression with nonlinear dimensionality reduction. Through a series of experiments and comparative analysis with supervised NLDR methods for a lot of typical test data the new proposed method is validated to outperform other supervised NLDR ones.