Learning a Lot from Only a Little: Genetic Programming for Panel Segmentation on Sparse Sensory Evaluation Data