Applying a hybrid IGA-SimE algorithm to a multimedia retrieval system

In this study, we propose a multimedia retrieval system framework using the Kansei retrieval agent (KRA) model to retrieve multimedia data. Our previous study involved recommender systems using the KRA model with reciprocal evaluation of each KRA and a hidden KRA model to optimize agent parameters; however, our previous method uses an interactive genetic algorithm (IGA) to optimize these parameters. Therefore, the performance and runtimes decreases with larger inputs. Then, in the present study, we use a hybrid IGA and simulated evolution (SimE) algorithm to improve optimization performance. Moreover, our proposed method employs rearrangement evaluation of multimedia data to reduce the evaluation load of users. Furthermore, we have verified the effectiveness of the proposed method via a numerical simulation using a pseudo-user that imitates user preferences. Our simulation results show that optimization performance of the proposed method is substantially higher than that of our previous method.