Backwarding: An Overfitting Control for Genetic Programming in a Remote Sensing Application

by: Denis Robilliard, Cyril Fonlupt

Artificial Evolution (2002), pp. 57-74.

View FullText article

Reviews [Write a review of this article]

There are no reviews of this article

Find related articles from these CiteULike users

Phanix

Find related articles with these CiteULike tags

genetic-algorithm, genetic-programming, overfitting

Abstract

Overfitting the training data is a common problem in supervised machine learning. When dealing with a remote sensing inverse problem, the PAR, overfitting prevents GP evolved models to be successfully applied to real data. We propose to use a classic method of overfitting control by the way of a validation set. This allows to go backward in the evolution process in order to retrieve previous, not yet overfitted models. Although this “backwarding” method performs well on academic benchmarks, there is not enough improvement to deal with the PAR. A new backwarding criterion is then derived using real satellite data and the knowledge of plausible physical bounds for the PAR coefficient in the geographical area that is monitored. This leads to satisfactory GP models and drastically improved images.

@incollection{citeulike:2616685, abstract = {Overfitting the training data is a common problem in supervised machine learning. When dealing with a remote sensing inverse problem, the PAR, overfitting prevents GP evolved models to be successfully applied to real data. We propose to use a classic method of overfitting control by the way of a validation set. This allows to go backward in the evolution process in order to retrieve previous, not yet overfitted models. Although this “backwarding” method performs well on academic benchmarks, there is not enough improvement to deal with the PAR. A new backwarding criterion is then derived using real satellite data and the knowledge of plausible physical bounds for the PAR coefficient in the geographical area that is monitored. This leads to satisfactory GP models and drastically improved images.}, author = {Robilliard, Denis and Fonlupt, Cyril }, citeulike-article-id = {2616685}, doi = {http://dx.doi.org/10.1007/3-540-46033-0\_20}, journal = {Artificial Evolution}, keywords = {genetic-algorithm, genetic-programming, overfitting}, pages = {57--74}, posted-at = {2008-03-31 16:56:14}, priority = {2}, title = {Backwarding: An Overfitting Control for Genetic Programming in a Remote Sensing Application}, url = {http://dx.doi.org/10.1007/3-540-46033-0\_20}, year = {2002} }

RIS record

TY - CHAP ID - citeulike:2616685 L3 - citeulike-article-id:2616685 TI - Backwarding: An Overfitting Control for Genetic Programming in a Remote Sensing Application JF - Artificial Evolution SP - 57 EP - 74 N2 - Overfitting the training data is a common problem in supervised machine learning. When dealing with a remote sensing inverse problem, the PAR, overfitting prevents GP evolved models to be successfully applied to real data. We propose to use a classic method of overfitting control by the way of a validation set. This allows to go backward in the evolution process in order to retrieve previous, not yet overfitted models. Although this “backwarding” method performs well on academic benchmarks, there is not enough improvement to deal with the PAR. A new backwarding criterion is then derived using real satellite data and the knowledge of plausible physical bounds for the PAR coefficient in the geographical area that is monitored. This leads to satisfactory GP models and drastically improved images. KW - genetic-algorithm KW - genetic-programming KW - overfitting AU - Robilliard, Denis AU - Fonlupt, Cyril PY - 2002/// UR - http://dx.doi.org/10.1007/3-540-46033-0_20 ER -

RIS BibTeX