An Overview of Extreme Learning Machine

Abstract

Extreme Learning Machine (ELM), as a new learning framework of Single Hidden Layer Feedforward Neural Network (SLFN), has become one of the hottest research directions in the field of artificial intelligence in recent years. It has been widely used in multiclass classification, human action recognition and other fields. ELM provides an efficient and unified learning framework for regression, classification, feature learning, and clustering. At the same time, ELM theories and algorithms have been improved for specific applications. This paper aims to provide a comprehensive review of existing research related to ELM. We first give an overview of the standard ELM. Then we discuss and analyze the typical variants of ELM, which are improved from different aspects, including models, strengths and weaknesses. Then we compare ELM with traditional methods in classification field and list the performance comparison between ELM and other deep network algorithms. Furthermore, the latest applications of ELM and its variants are summarized. Last but not least, we conclude the research trends of ELM variants mentioned above.