A study on lightweight Extreme Learning Machine algorithm for edge-computing

In recent years, AI technology has made remarkable progress and has become indispensable for us, such as autonomous car, medical image diagnostic, and speech translation. However, processing task of learning and inference tend to increase for these machine learning algorithm. High-performance computer will be needed to keep our daily life. On the other hand, edge-computing technologies have been implementing several devices, such as smartphone and sensor node. These devices also will be required to ensure real-time performance through high-speed processing. Therefore, we focus on an Extreme Learning Machine, which is a kind of compact machine learning algorithm. This paper presents the lightweight Extreme Learning Machine algorithm for edge-computing. We have achieved a correct classification rate of about 80 % for 3-value classification when trained on multiple data sets, which is sufficiently accurate.