Benchmark problems for development
- When developing novel GP methods, it is useful to compare new techniques
with old ones on a suite of benchmark problems that are easily scalable in
difficulty. The even-parity problem seems to be the most used such
benchmark, but results obtained only on that problem are not necessarily
Even parity seems popular because:
1. It is quite easy to compute fitness.
2. It is hard for GP without ADFs.
3. It is easy to scale by just increasing the number of variables.
Can anyone suggest problems with similar properties that could be useful
as test cases during development of new GP systems?
Department of Computer Science