Comparison of adaptive methods for function estimation from samples

Vladimir Cherkassky, Don Gehring, Filip Mulier

Research output: Contribution to journalReview articlepeer-review

155 Scopus citations


The problem of estimating an unknown function from a finite number of noisy data points has fundamental importance for many applications. This problem has been studied in statistics, applied math, engineering, artificial intelligence, and, more recently, in the fields of artificial neural networks, fuzzy systems, and genetic optimization. In spite of many papers describing individual methods, very little is known about the comparative predictive (generalization) performance of various methods. We discuss subjective and objective factors contributing to the difficult problem of meaningful comparisons. We also describe a pragmatic framework for comparisons between various methods, and present a detailed comparison study comprising several thousand individual experiments. Our approach to comparisons is biased toward general (nonexpert) users who do not have detailed knowledge of the methods used. Our study uses six representative methods described using a common taxonomy. Comparisons performed on artificial data sets provide some insights on applicability of various methods. No single method proved to be the best, since a method's performance depends significantly on the type of the target function (being estimated), and on the properties of training data (i.e., the number of samples, amount of noise, etc.). Hence, our conclusions contradict many known comparison studies (performed by experts) that usually show performance superiority of a single method (promoted by experts). We also observed the difference in a method's robustness, i.e., the variation in predictive performance caused by the (small) changes in the training data. In particular, statistical methods using greedy (and fast) optimization procedures tend to be less robust than neural-network methods using iterative (slow) optimization for parameter (weight) estimation.

Original languageEnglish (US)
Pages (from-to)969-984
Number of pages16
JournalIEEE Transactions on Neural Networks
Issue number4
StatePublished - 1996

Bibliographical note

Funding Information:
Manuscript received October 6, 1994; revised May 28, 1995 and January 4, 1996. This work was supported in part by the 3M Corporation. The authors are with the Department of Electrical Engineering, University of Minnesota, Minneapolis, MN 55455 USA. Publisher Item Identifier S 1045-9227(96)04398-6.


Dive into the research topics of 'Comparison of adaptive methods for function estimation from samples'. Together they form a unique fingerprint.

Cite this