An Evaluation of Support Vector Machines as a Pattern Recognition Tool

The purpose of this report is in examining the generalization performance of Support Vector Machines (SVM) as a tool for pattern recognition and object classification. The work is motivated by the growing popularity of the method that is claimed to guarantee a good generalization performance for the task in hand. The method is implemented in MATLAB. SVMs based on various kernels are tested for classifying data from various domains.