Analysis of schema variance and short term extinction likelihoods.pdf

Appeared in the Proceedings of Genetic Programming’98, Madison, Wisconsin Analysis of Schema Variance and Short Term Extinction Likelihoods Riccardo Poli W.B. Langdon Una-May O’Reilly School of Computer Science School of Computer Science Artificial Intelligence Lab The University of Birmingham The University of Birmingham MIT Edgbaston Edgbaston 545 Tech Sq Birmingham, B15 2TT, UK Birmingham, B15 2TT, UK Cambridge, MA, 02143 R.Poli@cs.bham.ac.uk W.B.Langdon@cs.bham.ac.uk unamay@ai.mit.edu Phone: +44-121-414-3739 Phone: +44-121-414-4791 Phone: +1-617-258-5978 ABSTRACT Schema analysis can also be combined with experimentation to yield a substantiated and further improved account of GP This paper first analyses the impact of [Poli and Langdon, 1997a]. variance on schema transmission. Work- Schema analysis does not lead to one “correct” GP schema ing from an exact derivation of the ex- theorem. Different GP schema definitions and formula- pected variance in schema transmission, tions of schema theorems should not be seen as competi- it derives and analyses the signal-to-noise tive but should be intellectually organised so as to distin-