Content deleted Content added
ref |
lx |
||
Line 1:
'''Interactive evolutionary computation''' (IEC) or '''aesthetic selection''' is a general term for methods of [[evolutionary computation]] that use human evaluation. Usually human evaluation is necessary when the form of [[fitness function]] is not known (for example, visual appeal or attractiveness; as in Dawkins, 1986<ref>{{cite book|author=Dawkins, R.|title=The Blind Watchmaker|publisher=Longman|year=1986|url=https://en.wikipedia.org/wiki/The_Blind_Watchmaker}}</ref>) or the result of optimization should fit a particular user preference (for example, taste of coffee or color set of the user interface).
==IEC design issues==
The number of evaluations that IEC can receive from one human user is limited by user fatigue which was reported by many researchers as a major problem. In addition, human evaluations are slow and expensive as compared to fitness function computation. Hence, one-user IEC methods should be designed to converge using a small number of evaluations, which necessarily implies very small populations. Several methods were proposed by researchers to speed up convergence, like interactive constrain evolutionary search (user intervention) or fitting user preferences using a [[convex function]]<ref>{{cite journal|author=Takagi, H.|year=2001|title=Interactive Evolutionary Computation: Fusion of the Capacities of EC Optimization and Human Evaluation|journal=Proceedings of the IEEE 89|issue=9|pages=1275-1296|url=http://www.design.kyushu-u.ac.jp/~takagi/TAKAGI/IECpaper/ProcIEEE_3.pdf|doi=10.1109/5.949485}}</ref>. IEC [[human-computer interface]]s should be carefully designed in order to reduce user fatigue.
However IEC implementations that can concurrently accept evaluations from many users overcome the limitations described above. An example of this approach is an interactive media installation by [[Karl Sims]] that allows to accept preference from many visitors by using floor sensors to evolve attractive 3D animated forms. Some of these multi-user IEC implementations serve as collaboration tools, for example [[HBGA]].
| |||