- author = "Thomas Loveard and Vic Ciesielski",
- title = "Genetic Programming for Classification: An Analysis of Convergence Behaviour",
- volume = "2557",
- pages = "309--320",
- year = "2002",
- CODEN = "LNCSD9",
- ISSN = "0302-9743",
- bibdate = "Sat Nov 30 20:58:15 MST 2002",
- acknowledgement = ack-nhfb,
- booktitle = "AI 2002: Advances in Artificial Intelligence : 15th Australian Joint Conference on Artificial Intelligence",
- editor = "Bob McKay and John Slaney",
- series = "Lecture Notes in Computer Science",
- address = "Canberra, Australia",
- month = "2-6 " # dec,
- publisher = "Springer",
- keywords = "genetic algorithms, genetic programming",
- ISBN = "3-540-00197-2",
-
URL = "
http://goanna.cs.rmit.edu.au/~vc/papers/aust-ai02.pdf",
-
DOI = "
doi:10.1007/3-540-36187-1_27",
-
abstract = "This paper investigates the unexpected convergence
behaviour of genetic Programming (GP) for
classification problems. Firstly the paper investigates
the relationship between computational effort and
attainable classification accuracy. Secondly we attempt
to understand why GP classifiers sometimes fail to
reach satisfactory levels of accuracy for certain
problems regardless of computational effort. The
investigation uses an artificially generated dataset
for which certain properties are known in advance for
the exploration of these areas. Results from this
artificial problem show that by increasing
computational effort, in the form of larger population
sizes and more generations, the probability of success
for a run does improve, but that the computational cost
far outweighs the rate of this success. Also, some
runs, even with very large populations running for many
generations, became stagnant and were unable to find an
acceptable solution. These results are also reflected
in real world classification problems.
From analysis of sub-tree components making up successful and unsuccessful programs it was noted that a small number of particular components were almost always present in successful programs, and that these components were often absent from unsuccessful programs. Also a variety of components appeared in unsuccessful programs that were never present in successful ones. Evidence from runs suggests that these components represent paths leading to optimal and sub-optimal branches in the evolutionary search space. Additionally, results suggest that if sub-optimal components (which mirror the concept of deception in genetic algorithms) are relatively greater in number than the optimal components for the problem, then the chances of GP finding a successful solution are reduced.",
