Article Preview
Top1. Introduction
In the area of product design, there is movement to add human feelings on the products for emphasizing value of the product. Along with the movement, engineering researches have been investigating human feelings from various perspectives. However, it is still difficult to model human feelings as an engineering model, therefore, it is still unsolved problem that making products suited to human feelings.
This problem is related to optimization: a product is represented as parameters, and solving the problem is considered as same as searching optimal solution of the parameters. Important thing in the search is how to evaluate the parameters constructing the product. Since human feelings are difficult to model and are like a black box, it is effective that solution candidates during the search are evaluated by human users subjectively. Interactive Evolutionary Computation (IEC) (Takagi, 2001) is a kind of Evolutionary Computation (EC), which is used for finding optimal solution in various problems. While EC is applied for computational and/or mathematical problems, IEC is generally applied for finding optimal or better solutions of media contents suited to each user’s feelings through interactions between human and computer system.
A representative example of IEC is an Interactive Genetic Algorithm (IGA) that derives the optimal or better solutions using Genetic Algorithm (GA). GA is an evolutionary algorithm that models the process of evolution of living organisms. IGA uses human feelings such as preference and impression as an evaluation function. The evaluation process is performed by scoring or selection by the human user.
According to Takagi’s survey (2001), IEC was applied for various optimization problems related to sense of sight, hearing, and smelling. Fundamental use of IEC is to find optimal or better solutions suited to each individual user. In terms of expanding the ability of IEC, some recent studies applied IEC into problem of multiple users. Previous studies proposed IEC reflecting multiple users’ feelings on computer graphics (Akase & Okada, 2014; Takenouchi, Tokumaru, & Muranaka, 2008; Takenouchi, Inoue, & Tokumaru, 2014; Sakai, Takenouchi, & Tokumaru, 2014; Ogawa, Miki, Hiroyasu, & Nagaya, 2001; Miki, Yamamoto, Wake, & Hiroyasu, 2006) and sounds (Fukumoto & Hatanaka, 2017; Nomura & Fukumoto, 2016). There were several methods for gathering evaluations of multiple users (described in a section 2.3. in detail). Miki et al. have proposed parallel Distributed IGA (DIGA) that is the extension of IGA in terms of using several islands where IGA processes are performed in each of the users in parallel (Ogawa, Miki, Hiroyasu, & Nagaya, 2001; Miki, Yamamoto, Wake, & Hiroyasu, 2006).
We have applied the DIGA for composing music melodies (Fukumoto & Hatanaka, 2017; Nomura & Fukumoto, 2016). In this method, each of multiple users performs IGA tasks respectively. Once we constructed the DIGA with synchronous model (Fukumoto & Hatanaka, 2017), after that, asynchronous model was also constructed (Nomura & Fukumoto, 2016). Purpose of this study is to investigate the efficiencies of the DIGA for composing music melody in comparison with general IGA without any exchange of solution candidate between the users. To investigate the efficiencies, listening experiment was conducted.