A semi-supervised learning method based on extreme learning machine
Min Han · Dalian Ligong Daxue xuebao · 2010
A semi-supervised learning method based on extreme learning machine(ELM)is proposed to solve two problems in semi-supervised learning,i.e.the slow learning speed and the increasing uncertainty.In the proposed method,firstly,ELM is extended from supervised learning to semisupervised learning.Secondly,the output threshold vector is utilized to control the extensibility degree of label samples.Finally,the alternate detection strategy is employed to evaluate the influence of uncertainty in extended samples.It is indicated by simulation results that the semi-supervised learning speed is significantly improved and the dependence for labeled samples is effectively reduced.