Next: 5 Scalability and Fitness
Up: 3 Symbolic Regression
Previous: 2 Regression Fitness Assessment
Many studies have used the Regression problem domain as a testbed.
The problem has been characterised by both Daida et al. (2001)
and O'Reilly (1998)
being unique in that it contains many conflicts of context and content
in the representation.
Keijzer (2003) found improvements
using interval arithmetic
and a simple fitting method, and other
improvements were found using more advanced models
for symbolic regression of polynomials [Nikolaev and Iba, 2001].