Penalties

[RomanKalkreuth]

High-level implementations of commonly used penalties that can be used for each GP module.

[domsob]

Penalties could, e.g., be controlled via the return value of the predict() method of a GP module. For example, returning float('nan') could trigger a predefined penalty. Using specific flags, the penalty could also be controlled to trigger different penalties depending on the context. The penalties could possibly be integrated into cost() in problem.py.