A new hybrid expansion function based mutual information for a multilayer neural networks optimization

الملخص EN

Function expansion was used to expand initial features based on a nonlinear transformation.

Many known expansion functions are found suchthe trigonometric, the polynomial, the Legendre polynomial, the powerseries, the exponential and the logarithmic transformation.

This paperpresent a comparison between different expansion functions based onmutual information and different performance functions.

We propose anew expansion process able to improve the correspondent mutualinformation and the final performance.

The process was tested; usingdifferent benchmark databases, and shows his ability to improve results ofclassification problems