On Improving Generalisation in Genetic Programming

Costelloe, Dan; Ryan, Conor

doi:10.1007/978-3-642-01181-8_6

Cited by 28 publications

(22 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Feature generation is the process of deriving new features from existing features (Guo, Jack, and Nandi 2005). In this technique, an evolutionary algorithm is used to generate and combine results of multiple independently discovered expression, e.g., by using a linear combination of GP results (Keijzer 2004;Costelloe and Ryan 2009), or by using non-linear function estimators applied to GE (de Silva, Noorian, Davis, and Leong 2013). This can be considered a type of machine learning and symbolic regression hybrid, as the final learning model is constructed from combination of simpler features created through a process similar to symbolic regression.…”

Section: Symbolic Regression and Feature Generationmentioning

confidence: 99%

gramEvol: Grammatical Evolution inR

Noorian¹,

Silva²,

Leong³

2016

J. Stat. Soft.

View full text Add to dashboard Cite

We describe an R package which implements grammatical evolution (GE) for automatic program generation. By performing an unconstrained optimization over a population of R expressions generated via a user-defined grammar, programs which achieve a desired goal can be discovered. The package facilitates the coding and execution of GE programs, and supports parallel execution. In addition, three applications of GE in statistics and machine learning, including hyper-parameter optimization, classification and feature generation are studied.

show abstract

Section: Symbolic Regression and Feature Generationmentioning

confidence: 99%

gramEvol: Grammatical Evolution inR

Noorian¹,

Silva²,

Leong³

2016

J. Stat. Soft.

View full text Add to dashboard Cite

show abstract

“…Costelloe and Ryan [1] experimentally concluded that GP with linear scaling may perform better compared to standard GP on training data, but the technique does not generalize well on test data. They proposed to combine "No Same Mate" selection with linear scaling to improve generalization ability of evolved GP solutions.…”

Section: F "Linear Scaling" With "No Same Mate" Selectionmentioning

confidence: 99%

Empirical modeling using genetic programming: a survey of issues and approaches

Dabhi

Chaudhary

2014

Nat Comput

View full text Add to dashboard Cite

Abstract--In the field of empirical modeling using Genetic Programming (GP), it is important to evolve solution with good generalization ability. Generalization ability of GP solutions get affected by two important issues: bloat and over-fitting. We surveyed and classified existing literature related to different techniques used by GP research community to deal with these issues. We also point out limitation of these techniques, if any. Moreover, the classification of different bloat control approaches and measures for bloat and over-fitting are also discussed. We believe that this work will be useful to GP practitioners in following ways: (i) to better understand concepts of generalization in GP (ii) comparing existing bloat and over-fitting control techniques and (iii) selecting appropriate approach to improve generalization ability of GP evolved solutions.

show abstract

“…For example, the "Quartic Polynomial" problem (using the function f (x) = x 4 +x 3 +x 2 +x) is probably the most used benchmark in GP [15].…”

Section: A Symbolic Regressionmentioning

confidence: 99%