Single-Hidden Layer Feedforward Neual Network Training Using Class Geometric Information

摘要

Single-hidden Layer Feedforward (SLFN) networks have been proven to be effective in many pattern classification problems. In this chapter, we provide an overview of a relatively new approach for SLFN network training that is based on Extreme Learning. Subsequently, extended versions of the Extreme Learning Machine algorithm that exploit local class data geometric information in the optimization process followed for the calculation of the network output weights are discussed. An experimental study comparing the two approaches on facial image classification problems concludes this chapter.