An efficiency of interactive differential evolution for optimization of warning sound with reflecting individual preference

Abstract

Interactive evolutionary computation (IEC) is a method for creating media contents suited to the user's preference. Efficiency of previous IEC studies was generally investigated from the view point of optimization. However, few previous studies paid attention to their efficiency as a reflection of individual preferences. As an application of IEC, we already proposed a method that creates sign sounds using paired comparison-based interactive differential evolution (IDE). This study investigates the efficiency of the IDE for reflecting individual difference on the created sound through two experiments. In Experiment 1, the subjects searched better warning sounds with the IDE. The result of Experiment 1 showed gradual convergence of DE's vectors. In Experiment 2, the subjects evaluated four sign sounds created in Experiment 1. Two of them were created by the subject, and the other two were created by another subject. The result of Experiment 2 showed significant increase in the subjective fitness value for sounds created by the subjects themselves. Furthermore, significant difference between sounds by one subject and another was observed in the final generation. These results suggest that IDE is effective in creating warning sounds suited to each user reflecting individual difference.

A Proposal for Continuous Evaluation-based Interactive Evolutionary Computation

“…These music notes also play roles of extension of length of music note and rest note. In the previous study, the IDE successfully obtained user's "warning" sign sound throughout 13 generations [10]. …”

“…In this study, the efficiency of the proposed IEC is investigated by using an IEC system creating sign sounds. Some previous studies proposed IEC that searches sign sounds by employing Genetic Algorithm and Differential Evolution (DE) [9,10].…”

“…This study uses Interactive Differential Evolution (IDE) creating sign sounds [10]. IDE is an interactive type of DE which proceeds its search with a comparison of the current solution (target vector) and the new solution (trial vector) [11,12].…”

“…The number of solution candidates (DE vectors) was eight in one generation. DE's parameters were set as F=0.9 and CR= 0.6 respectively by referring to the previous study [10,15,16].…”

Investigation of the efficiency of continuous evaluation‐based interactive evolutionary computation for composing melody

“…This study uses a system based on interactive differential evolution (IDE) creating sign sounds . The system applies paired comparison‐based IDE .…”