BASELINE GENETIC PROGRAMMING: SYMBOLIC REGRESSION ON BENCHMARKS FOR SENSORY EVALUATION MODELING

We introduce hedonic, modeling benchmarks for the field of sensory science evaluation. Our benchmark framework provides a general means of defining a response surface which we call a "sensory map". A sensory map is described by a mathematical expression which rationalizes domain specific knowledge of the explanatory variables and their individual or higher order contribution to hedonic, response. The benchmark framework supports the sensory map's socalled ground truth to be controllably distorted to mimic the human and protocol factors that obscure it. To provide a baseline for future algorithm comparison, we evaluate a public research release of genetic programming symbolic regression algorithm on a sampling of the framework's benchmarks.