Unknown function approximation, given input-output measurements, using Genetic Algorithm
Preliminary implementation of a Genetic Algorithm, utilized to approximate an unknown function, given input-output measurements. Flexibility is provided to obtain different results and bias the approximation towards a desired direction, through selection of multiple parameters such as, the size of the population, the k more accurate chromosomes to pick, the number of generations, and the tester-set of the input values.
note
For optimal approximation, the threshold (desired MSE) must be kept low enough, anywhere in the range (10^-5, 10^-7). This however, translates to excessive computational requirements, and a pretty poor performance.