Preventing overfitting in GP with canary functions

@inproceedings{Foreman2005PreventingOI,
  title={Preventing overfitting in GP with canary functions},
  author={Nate Foreman and Matthew P. Evett},
  booktitle={GECCO},
  year={2005}
}
Overfitting is a fundamental problem of most machine learning techniques, including genetic programming (GP). Canary functions have been introduced in the literature as a concept for preventing overfitting by automatically recognizing when it starts to occur. This paper presents a simple scheme for implementing canary functions using cross-validation. The… CONTINUE READING