Robust extreme learning machine for regression problems with its application to wifi based indoor positioning system

We propose two kinds of robust extreme learning machines (RELMs) based on the close-to-mean constraint and the small-residual constraint respectively to solve the problem of noisy measurements in indoor positioning systems (IPSs). We formulate both RELMs as second order cone programming problems. The fact that feature mapping in ELM is known to users is exploited to give the needed information for robust constraints. Real-world indoor localization experimental results show that, the proposed algorithms can not only improve the accuracy and repeatability, but also reduce the deviations and worst case errors of IPSs compared with basic ELM and OPT-ELM based IPSs.