A study on induction of support vector machines for portable devices

In recent years, portable devices like IC cards and smart phones are becoming more and more popular. To use these kinds of devices more efficiently, it is desired to embed some intelligent agent (IA) in them. Support vector machine (SVM) is a kind of IA that can help a human user equipped with portable devices to be aware of different situations and make proper decisions. One problem in using SVM in portable devices is that the implementation cost is usually high. In our research, we have tried to reduced this cost through dimensionality reduction (DR). Since the implementation cost of the DR part should also be considered in a restricted computing platform, we adopted a simple centroid based DR. In this paper, we conducted a large number of experiments, and show that compact SVMs can be obtained through DR, but the performance of the system needs to be improved.