Short Paper: Object Oriented Machine Learning with a Multicore Real-Time Java Processor

摘要

The term intelligent systems is spreading beyond the data mining and machine learning communities. This presents new challenges that are fundamental to classical problems within object oriented programming and analysis. In this paper we investigate the use of a popular intelligent algorithm on a Java-based processor. The processor is a realtime enabled processor implemented on an FPGA, and we deploy a support vector machine on this processor. Furthermore, we show how this support vector machine can work on the Java-processor's multiple cores. This is a first step toward understanding how intelligent algorithms can be implemented on object-oriented Java systems with multiple cores in a hard real-time environment. Our experiments show significant speedup of the selected machine learning algorithm, and this can potentially be useful for other intelligent algorithms also.