User's favorite scent design using paired comparison-based Interactive Differential Evolution

This study proposes a method that creates a scent suited with a user's favor using paired comparison-based Interactive Differential Evolution. In the proposed method, the user smells two scents and selects the preferred one by simple comparison. Based on the repetitive comparisons, Differential Evolution (DE) optimizes the scent suited with the user. Each scent is composed of several source scents, and strength of each source scent is described as values in DE's vector. To investigate the efficacy of the proposed method fundamentally, smelling experiments composed of comparing experiment and evaluating experiment are performed. In the comparing experiment, the subjects compare presented pair scents and select the preferred one through ten generations, and DE evolves scent to user's favor based on the comparisons. In the evaluating experiment, the subjects evaluate four representative scents picked from 0-, 3-, 6-, 9-th generations, respectively. The results of the experiments showed a tendency of the increase of fitness value in accordance with evolution.