%0 Conference Paper %1 ieee94:kinnear %A Kinnear, Jr., Kenneth E. %B Proceedings of the 1994 IEEE World Conference on Computational Intelligence %C Orlando, Florida, USA %D 1994 %I IEEE Press %K (artificial adaptive algorithm algorithms, autocorrelation, basin depths, fitness genetic intelligence), landscape landscapes, learning measures, problems, programming, random search theory, walks walks, %P 142--147 %R doi:10.1109/ICEC.1994.350026 %T Fitness Landscapes and Difficulty in Genetic Programming %U http://ieeexplore.ieee.org/iel2/1125/8059/00350026.pdf?isNumber=8059 %V 1 %X The structure of the fitness landscape on which genetic programming operates is examined. The landscapes of a range of problems of known difficulty are analyzed in an attempt to determine which landscape measures correlate with the difficulty of the problem. The autocorrelation of the fitness values of random walks, a measure which has been shown to be related to perceived difficulty using other techniques, is only a weak indicator of the difficulty as perceived by genetic programming. All of these problems show unusually low autocorrelation. Comparison of the range of landscape basin depths at the end of adaptive walks on the landscapes shows good correlation with problem difficulty, over the entire range of problems examined. %@ 0-7803-1899-4

@inproceedings{ieee94:kinnear, abstract = {The structure of the fitness landscape on which genetic programming operates is examined. The landscapes of a range of problems of known difficulty are analyzed in an attempt to determine which landscape measures correlate with the difficulty of the problem. The autocorrelation of the fitness values of random walks, a measure which has been shown to be related to perceived difficulty using other techniques, is only a weak indicator of the difficulty as perceived by genetic programming. All of these problems show unusually low autocorrelation. Comparison of the range of landscape basin depths at the end of adaptive walks on the landscapes shows good correlation with problem difficulty, over the entire range of problems examined.}, added-at = {2008-06-19T17:35:00.000+0200}, address = {Orlando, Florida, USA}, author = {{Kinnear, Jr.}, Kenneth E.}, biburl = {https://www.bibsonomy.org/bibtex/25fb2ccc9db4c38e14dbd5cf676bb278c/brazovayeye}, booktitle = {Proceedings of the 1994 IEEE World Conference on Computational Intelligence}, doi = {doi:10.1109/ICEC.1994.350026}, interhash = {5f0f93e9a2332f2f6bc2f10822666983}, intrahash = {5fb2ccc9db4c38e14dbd5cf676bb278c}, isbn = {0-7803-1899-4}, keywords = {(artificial adaptive algorithm algorithms, autocorrelation, basin depths, fitness genetic intelligence), landscape landscapes, learning measures, problems, programming, random search theory, walks walks,}, month = {27-29 June}, notes = {Defines difficulty as number of fitness cases/1000. Considers a few parity and sort problems. Fitness landscape investigated by using GP operators (without selection) on gen=0 to give a number of random walks. Look at autocorrelation of fitness along these walks. Essentially none (<0.5) very much worse than published GA. Also little correlation between this and difficulty measure.}, pages = {142--147}, publisher = {IEEE Press}, size = {6 pages}, timestamp = {2008-06-19T17:43:20.000+0200}, title = {Fitness Landscapes and Difficulty in Genetic Programming}, url = {http://ieeexplore.ieee.org/iel2/1125/8059/00350026.pdf?isNumber=8059}, volume = 1, year = 1994 }