- author = "W. B. Langdon and R. Poli",
- title = "Better Trained Ants for Genetic Programming",
- institution = "University of Birmingham, School of Computer Science",
- number = "CSRP-98-12",
- month = apr,
- year = "1998",
- keywords = "genetic algorithms, genetic programming, multi-objective GP",
- file = "/1998/CSRP-98-12.ps.gz",
-
URL = "
ftp://ftp.cs.bham.ac.uk/pub/tech-reports/1998/CSRP-98-12.ps.gz",
-
abstract = "The problem of programming an artificial ant to follow
the Santa Fe trail has been repeatedly used as a
benchmark problem in GP. Recently we have shown
performance of several techniques is not much better
than the best performance obtainable using uniform
random search. We suggested that this could be because
the program fitness landscape is difficult for hill
climbers and the problem is also difficult for Genetic
Algorithms as it contains multiple levels of
deception.
Here we redefine the problem so the ant is (1) obliged to traverse the trail in approximately the correct order, (2) to find food quickly. We also investigate (3) including the ant's speed in the fitness function, either as a linear addition or as a second objective in a multi-objective fitness function, and (4) GP one point crossover.
A simple genetic programming system, with no size or depth restriction, is shown to perform approximately three times better with the improved training function. (Extends CSRP-98-08 \cite{langdon:1998:antlook})",
