The impact of transformation function on the classification ability of complex valued extreme learning machines

Classification is a rather omnipresent problem in many of the technological areas ranging from image processing to medical applications. With complex-valued neural network classifiers posing better decision making capabilities due to its orthogonal decision boundaries and it's comparatively better computational capability many complex valued neural network (CVNN) classifiers has been presented in literature. In this paper a review on the state of the art on a family of CVNNs known as complex valued extreme learning machines (CELM) is presented. With their better generalization ability and lesser computational efforts for classification problems CELMs provide a better solution for real-valued classification problems. The four CELMs that is used for solving real valued classification problems namely, Circular CELM (CC-ELM), Phase encoded CELM (PE-CELM), Bilinear Branch cut CELM (BB-CELM) and Fast Learning Complex valued Neural Classifier (FLCNC). The evaluations are done based on the datasets available in the UCI repository. Through this study it could proved that the synergy between the ELM and CVNN has brought better results in the classification arena.