Performance of different models in non-linear subspace
Jyostna Devi Bodapati, Naralasetti Veeranjaneyulu · 2016
The problem of high dimensionality has been gaining increased focus in the recent literature on pattern recognition. This is due to the increase in the availability of the high volumes of data in various fields. Multiple sensors are being used to extract the data and each sensor gives multiple features. On the other side performance of a classifier depends on the type of features that are used to represent the data. But storing the high dimensional data is a challenge as they occupy more space and also demands more computational resources. In generative models like Gaussian Mixture Model(GMM) based classification there is a direct correlation between the number of features used to represent the data and the number of parameters that are to be estimated. This is where the dimensionality reduction would be helpful. This paper aims to demonstrate the performance of different classifiers in the reduced subspace. Based on the literature it has been observed that non-linear projection is helpful for classification than linear projection of the data. The major objective of this paper is to show the performance of different classifiers in the reduced space after linear and non-linear projection of the data. To reduce dimensionality PCA, KPCA and Auto encoder based techniques are used and to compare the performance in original and reduced space, Classification models like Gaussian Mixture Models (GMM), Artificial Neural Network (ANN) and Support vector machine (SVM) based classifiers are used.