NSF-supported agricultural engineers from the University of Arkansas have demonstrated a new technique for training machine-learning systems to incorporate human perception when predicting food quality.