Incremental Learning Algorithms for Classification and Regression: local strategies

Florence d'Alché-Buc 1 Liva Ralaivola 1
1 APA - Apprentissage et Acquisition des connaissances
LIP6 - Laboratoire d'Informatique de Paris 6
Abstract : We present a new local strategy to solve incremental learning tasks. It allows to avoid re‐learning of all the parameters by selecting a working subset where the incremental learning is performed. While this procedure can be applied to various schemes (hybrid decision trees, committee machines), we illustrate it with Support Vector Machines based on local kernel. We derive and compare three methods to perform the selection procedure: two of them take advantage of the estimation of generalization error by using theoretical error bounds devoted to SVM. Experimental simulations on three typical datasets of machine learning give promising results.
Document type :
Conference papers
Complete list of metadatas

https://hal.archives-ouvertes.fr/hal-01570821
Contributor : Lip6 Publications <>
Submitted on : Monday, July 31, 2017 - 5:05:39 PM
Last modification on : Thursday, March 21, 2019 - 1:12:48 PM

Identifiers

Citation

Florence d'Alché-Buc, Liva Ralaivola. Incremental Learning Algorithms for Classification and Regression: local strategies. CASYS 2001 - Fifth International Conference on COMPUTING ANTICIPATORY SYSTEMS, Aug 2001, Liege, Belgium. pp.320-329, ⟨10.1063/1.1503700⟩. ⟨hal-01570821⟩

Share

Metrics

Record views

56