Parzen Density Estimator for Complex Pattern Classification and High Dimensional Data Analysis
Chulhee Lee, Seongyoun Woo, Jaesung Rim · 2018
In this paper, a new method is proposed, which can be used to classify complex pattern classification problems with non-linear decision boundaries and to perform high dimensional data analysis. The proposed method first applies a linear feature extraction procedure and reduces the input dimension. The feature extraction is based on the assumption that a vector connecting two samples from different classes contains discriminantly useful information for classification. Then, the proposed method applies the Parzen density estimator in a reduced feature space. Experimental results show that the proposed method showed classification performance comparable to SVM and DNN. The proposed method can be also used to analyze high dimensional data such as feature spaces produced by DCN.