Exploring Overfitting in Genetic Programming

Paris, Grégory; Robilliard, Denis; Fonlupt, Cyril

doi:10.1007/978-3-540-24621-3_22

Cited by 32 publications

(17 citation statements)

References 4 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…There are different reasons why overfitting can occur. The existence of noise in training samples can cause a model to be fit to the data which is more complex than the true underlying model [14]. For symbolic regression, an example would be fitting a high order polynomial to noisy data, which happens to pass through all training points, when the true function is in fact a lower order polynomial.…”

Section: Introductionmentioning

confidence: 99%

A Preliminary Investigation of Overfitting in Evolutionary Driven Model Induction: Implications for Financial Modelling

Tuite

Agapitos

O’Neill

et al. 2011

Applications of Evolutionary Computation

View full text Add to dashboard Cite

Abstract. This paper investigates the effects of early stopping as a method to counteract overfitting in evolutionary data modelling using Genetic Programming. Early stopping has been proposed as a method to avoid model overtraining, which has been shown to lead to a significant degradation of out-of-sample performance. If we assume some sort of performance metric maximisation, the most widely used early training stopping criterion is the moment within the learning process that an unbiased estimate of the performance of the model begins to decrease after a strictly monotonic increase through the earlier learning iterations. We are conducting an initial investigation on the effects of early stopping in the performance of Genetic Programming in symbolic regression and financial modelling. Empirical results suggest that early stopping using the above criterion increases the extrapolation abilities of symbolic regression models, but is by no means the optimal training-stopping criterion in the case of a real-world financial dataset.

show abstract

Section: Introductionmentioning

confidence: 99%

A Preliminary Investigation of Overfitting in Evolutionary Driven Model Induction: Implications for Financial Modelling

Tuite

Agapitos

O’Neill

et al. 2011

Applications of Evolutionary Computation

View full text Add to dashboard Cite

show abstract

“…For example, the existence of noise in training samples can cause a model to be fit to the data which is more complex than the true underlying model [19]. For symbolic regression, an example would be fitting a high order polynomial to noisy data, which happens to pass through all training points, when the true function is in fact a lower order polynomial.…”

Section: Causes Of Overfittingmentioning

confidence: 99%

“…The more training data available, the more likely we are to discover the true underlying model, and the less likely we are to settle on a spurious result. Overfitting is also more likely to occur with complex hypotheses [19]. Learning algorithms that are run for a long time are more likely to trigger overfitting, than if they had been run for a shorter time period [3].…”

Section: Causes Of Overfittingmentioning

confidence: 99%

See 1 more Smart Citation

Tackling Overfitting in Evolutionary-Driven Financial Model Induction

Tuite

Agapitos

O’Neill

et al. 2011

Natural Computing in Computational Finance

View full text Add to dashboard Cite

“…Overfitting is a problem which can arise in machine learning and optimisation techniques such as Genetic Programming (GP) [7,1]. A model is described as overfitting the data if, while having a good fit on the training data, there exists an alternative model which fits the data as a whole better, despite having a worse fit on the training data [4].…”

Section: Overfitting and Early Stoppingmentioning

confidence: 99%