Extreme Learning Machine (Neural Network) with Regularization Parameters optimized by a Genetic Algorithm. For my "Natural Computing" Class.
This package presents a GA-ELM model that uses a genetic algorithm (with the basic Octave GA Package) to find the Generalized Tikhonov Regularization process.
One should refer to the GA package on SourceForge to understanding its functions.
It's awkwardly suboptimal (the best error found was around 7.7%), but it may be a good starting point to other people.
I think that the GA package has bad crossover and mutation policies and the user should write his own code.