"Genetic Programming : Evolutionary Approaches to Multistrategy Learning", Hugo de Garis, ch. 21 in "Machine Learning : A Multistrategy Approach, Vol. 4", R.S. Michalski & G.Tecuci (Eds.), Morgan Kauffman, 1992.  Home/Search   Document Not in Database   Summary   Related Articles

....paper to pursue a third aim, namely to try to grow nonconvex artificial shapes of colonies of cells in reproductive cellular automata. Earlier publications aimed at creating an Artificial Embryology resulted in the growth of successful convex shapes but nonconvex shapes failed to evolve [e.g. de GARIS 1992b] The new field of Artificial Embryology (or at least the self assembly of complex systems) is felt to be important for the future development of complex system design. New technologies, such as WSI (Wafer Scale Integration) Molecular Electronics [CARTER et al. 1988] and Nanotechnology ....

....type of evolution (for triangular and rectangular target shapes) are shown in FIG. 4, with one frame per cycle or iteration. Note that NI evolved to be 4, for both shapes. Fitness = 93 ) Fitness = 95 ) FIG. 4 TRIANGULAR RECTANGULAR TARGET SHAPES The above ideas were tested on various shapes [de GARIS 1992b] both convex and non convex. The convex shapes worked well (with fitness values around 95 , e.g. FIG. 4) but non convex shapes evolved poorly, with low fitness values (e.g. as low as 20 for an arch shape) The challenge then was to evolve arbitrary non convex shapes, e.g. 3D embryo shapes ....

[Article contains additional citation context not shown here]

"Genetic Programming : Evolutionary Approaches to Multistrategy Learning", Hugo de Garis, ch. 21 in "Machine Learning : A Multistrategy Approach, Vol. 4", R.S. Michalski & G.Tecuci (Eds.), Morgan Kauffman, 1992.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC