On Using Prototype Reduction Schemes and Classifier Fusion Strategies to Optimize Kernel-Based Nonlinear Subspace Methods

摘要

In Kernel based Nonlinear Subspace (KNS) methods, the length of the projections onto the principal component directions in the feature space, is computed using a kernel matrix, K, whose dimension is equivalent to the number of sample data points. Clearly this is problematic, especially, for large data sets. To solve the problem, in [9] we earlier proposed a method of reducing the size of the kernel by invoking a Prototype Reduction Scheme (PRS) to reduce the data into a smaller representative subset, rather than define it in terms of the entire data set. In this paper we propose a new KNS classification method for further enhancing the efficiency and accuracy of the results presented in [9]. By sub-dividing the data into smaller subsets, we propose to employ a PRS as a pre-processing module, to yield more refined representative prototypes. Thereafter, a Classifier Fusion Strategies (CFS) is invoked as a post-processing module, so as to combine the individual KNS classification results to derive a consensus decision. Our experimental results demonstrate that the proposed mechanism significantly reduces the prototype extraction time as well as the computation time without sacrificing the classification accuracy. The results especially demonstrate that the computational advantage for large data sets is significant when a parallel programming philosophy is applied.