Abstract
How do we find objects in scenes? When we recognize things in the outside world, we feel that we recognize shapes as one entire object, not an ensemble of inconsistent characteristics. In perception theory, it is thought that features are integrated in structures that define objects, but how these structures are maintained in the working memory remains unknown. Therefore submitting clear evidence of the features binding in the memory is the first goal of this study. In this experiment, firstly the target is displayed. Then a panel is shown with the target and some distractors. Firstly, we generated eight different individual targets in this experiment. We show the targets in different presentation patterns: for each target, the presentation time is 300ms or 600ms, and the number of distractors can be 15 and 23 for a total of four different trial combinations. We measured the correct answer rate and the reaction time in those four tasks. The targets are generated with the using Genetic Algorithm. The targets more easy to remember will have a lower reaction time and better correct answer rate. Using these results, the genetic algorithm produces next generation targets in order to be more easy to remember. As a result, the “correct answer” rate is 51% at stimulation time of 300ms, 61% at 600ms in the first generation, and it is 79% at 300ms, 73% at 600ms in the second generation. In addition, we found that in the second generation there is a clear reduction of reaction times. When the number of distractors increases we found that the reaction time gets longer and this is preserved in the second generation too. This behavior is in accord with the visual search “conjunction search” theory. We will describe our Genetic Algorithm model able to generate targets easy to remember.