Interactive Population-Based Incremental Learning for Problems with Implicit Performance Indices

An interactive population-based incremental learning (IPBIL) algorithm has been proposed to optimize problems with implicit performance indices, which were traditionally solved by using interactive evolutionary computation (IEC). That is expected to reduce user fatigue, which is a key limitation of IEC, because users only need to select some good individuals rather than evaluate all individuals when using IPBIL. To compare the performance of IEC and IPBIL, they were applied to a fashion design system, a problem with implicit performance indices. Experimental results indicate that although IPBIL needs more generations to find a satisfactory design, it needs less time consumption and much fewer mouse clicks than IEC. Accordingly, compared with IEC, IPBIL can significantly reduce user fatigue.