Leave One Out Error, Stability, and Generalization of Voting Combinations of Classifiers

Authors:
Theodoros Evgeniou;Massimiliano Pontil;André Elisseeff
Affiliations:
Technology Management, INSEAD, Boulevard de Constance, 77305 Fontainebleau, France. theodoros.evgeniou@insead.edu;DII, University of Siena, Via Roma 56, 53100 Siena, Italy. pontil@dii.unisi.it;Max Planck Institute for Biological Cybernetics, Spemannstrasse 38, 72076 Tübingen, Germany. andre.elisseeff@tuebingen.mpg.de
Venue:
Machine Learning
Year:
2004

Citing 9
Cited 16

The nature of statistical learning theory

The nature of statistical learning theory
Bagging predictors

Machine Learning
Advances in kernel methods: support vector learning

Advances in kernel methods: support vector learning
Fast training of support vector machines using sequential minimal optimization

Advances in kernel methods
Algorithmic stability and sanity-check bounds for leave-one-out cross-validation

Neural Computation
Exploiting generative models in discriminative classifiers

Proceedings of the 1998 conference on Advances in neural information processing systems II
A sharp concentration inequality with application

Random Structures & Algorithms
Support Vector Machines: Training and Applications

Support Vector Machines: Training and Applications
Stability and generalization

The Journal of Machine Learning Research

Stability of Randomized Learning Algorithms

The Journal of Machine Learning Research
Bootstrapping rule induction to achieve rule stability and reduction

Journal of Intelligent Information Systems
Demonstrating the stability of support vector machines for classification

Signal Processing - Signal processing in UWB communications
Gene expression profile class prediction using linear Bayesian classifiers

Computers in Biology and Medicine
Constrained Local Regularized Transducer for Multi-Component Category Classification

PRICAI '08 Proceedings of the 10th Pacific Rim International Conference on Artificial Intelligence: Trends in Artificial Intelligence
Data dependency in multiple classifier systems

Pattern Recognition
EnsembleMatrix: interactive visualization to support machine learning with multiple classifiers

Proceedings of the SIGCHI Conference on Human Factors in Computing Systems
Ensembles of partially trained SWMs with multiplicative updates

IJCAI'07 Proceedings of the 20th international joint conference on Artifical intelligence
Interactive optimization for steering machine classification

Proceedings of the SIGCHI Conference on Human Factors in Computing Systems
An overproduce-and-choose strategy to create classifier ensembles with tuned SVM parameters applied to real-world fault diagnosis

CIARP'10 Proceedings of the 15th Iberoamerican congress conference on Progress in pattern recognition, image analysis, computer vision, and applications
Robust SVM-based biomarker selection with noisy mass spectrometric proteomic data

EuroGP'06 Proceedings of the 2006 international conference on Applications of Evolutionary Computing
A game-theoretic approach to weighted majority voting for combining SVM classifiers

ICANN'06 Proceedings of the 16th international conference on Artificial Neural Networks - Volume Part I
Bayesian learning with local support vector machines for cancer classification with gene expression data

EC'05 Proceedings of the 3rd European conference on Applications of Evolutionary Computing
Automated Diagnosis of Alzheimer Disease using the Scale-Invariant Feature Transforms in Magnetic Resonance Images

Journal of Medical Systems
GRADIENT: Grammar-driven genetic programming framework for building multi-component, hierarchical predictive systems

Expert Systems with Applications: An International Journal
Model combination for support vector regression via regularization path

PRICAI'12 Proceedings of the 12th Pacific Rim international conference on Trends in Artificial Intelligence

Quantified Score

Hi-index	0.00

Visualization

Abstract

We study the leave-one-out and generalization errors of voting combinations of learning machines. A special case considered is a variant of bagging. We analyze in detail combinations of kernel machines, such as support vector machines, and present theoretical estimates of their leave-one-out error. We also derive novel bounds on the stability of combinations of any classifiers. These bounds can be used to formally show that, for example, bagging increases the stability of unstable learning machines. We report experiments supporting the theoretical findings.