Evolving the stimulus to fit the brain: Investigating visual search in complex environments using genetic algorithms E Van der Burg1, J Cass2, J Theeuwes3, D Alais1 |
|---|
1School of Psychology, University of Sydney, Australia |
| Using a visual search display too complex to be tractable with conventional methods, we applied a genetic algorithm to investigate how observers search within complex visual environments. Starting with a population of random displays (136 distractors per display, varying in colour and orientation), the genetic algorithm mimics natural selection by combining over successive generations displays affording fastest search (the ‘fittest’) and discarding all others. For all observers, displays affording efficient search evolved very rapidly. From first-generation search times of ~5 s, search times declined rapidly over just 14 generations. Interestingly, all observers evolved similar displays even though the search space was large and the evolution unconstrained. Specifically, colour evolved first, followed by orientation. This pattern was not predicted by current models of visual search. The genetic algorithm, therefore permits highly efficient search of multidimensional spaces and produces consistent evolution patterns that point to the brain’s own search strategies and preferred saliency cues. This a-theoretical approach provides unique insights into complex visual search and is adaptable to a wide range of paradigms which, until now, have been intractable using traditional methods. |
| Up Home |
|---|