Abstract

Large memory space is needed for learning Support Vector Machine (SVM) and the speed of optimization becomes extremely slow with large-scale training set. Many Support Vectors brings difficulty in practical application of SVM. To solve these problems, this study presents a SVM based on the AiNet (Artificial Immune Network Model). The AiNet is used to compress the large-scale training set, to remove redundant data and then to classify different objects according to similarity. In this way, the training set is compressed to a comparatively small-scale training set which is used to implement the training and to build the final classification model. The practical application shows the compression of training set using the AiNet reduces the learning cost, improves the speed of the classification and obtains the classification accuracy better than classification model trained directly with large-scale sample set.

Paper Details