- author = "Gang Li and Kin Hong Lee and Kwong-Sak Leung",
- title = "Using Instruction Matrix Based Genetic Programming to Evolve Programs",
- booktitle = "Proceedings of the Second International Symposium on Computation and Intelligence, ISICA 2007",
- year = "2007",
- editor = "Lishan Kang and Yong Liu and Sanyou Y. Zeng",
- volume = "4683",
- series = "Lecture Notes in Computer Science",
- pages = "631--640",
- address = "Wuhan, China",
- month = sep # " 21-23",
- publisher = "Springer",
- keywords = "genetic algorithms, genetic programming, Instruction Matrix based Genetic Programming",
- isbn13 = "978-3-540-74580-8",
-
DOI = "
doi:10.1007/978-3-540-74581-5_69",
- size = "10 pages",
- abstract = "In Genetic Programming (GP), evolving tree nodes separately would be an ideal approach to reduce the huge solution space of GP. We use Instruction Matrix based Genetic Programming (IMGP) to evolve tree nodes separately while taking into account their interdependencies in the form of subtrees. IMGP uses an Instruction Matrix (IM) to maintain the statistical data of tree nodes and subtrees. IMGP extracts program trees from IM, and updates IM with the information of the extracted program trees. The experiments have verified that the results of IMGP are better than those the related GP algorithms in terms of the qualities of the solutions and the number of program evaluations.",
-
notes = "'no explicit population' cites \cite{shan:2004:gmpe}
\cite{Shan:2003:Pewel} assumes full binary tree
(hs-expression) so every part of tree has a known grid
position. (Limits three height)
'like context preserving crossover' \cite{Dhaeseleer:1994:cpcGP} IMGP matrix shuffle. 6-parity also with sin,cos,exp,rlog. Max. Symbolic regression like PIPE? Fitness of primitive by grid position (rather than by neighbours in sub-tree).",
- bibdate = "2007-08-31",
- bibsource = "DBLP, http://dblp.uni-trier.de/db/conf/isica/isica2007.html#LiLL07",
