Combining binary classifiers with imprecise probabilities

Authors:
Sébastien Destercke;Benjamin Quost
Affiliations:
INRA/CIRAD, UMR1208, Montpellier cedex, France;HEUDIASYC, UMR UTC-CNRS 6599. Université de Technologie de Compiègne. Centre de Recherches de Royallieu. Compiégne, France
Venue:
IUKM'11 Proceedings of the 2011 international conference on Integrated uncertainty in knowledge modelling and decision making
Year:
2011

Citing 10
Cited 1

The transferable belief model

Artificial Intelligence
Classification by pairwise coupling

NIPS '97 Proceedings of the 1997 conference on Advances in neural information processing systems 10
Approximate statistical tests for comparing supervised classification learning algorithms

Neural Computation
Probability Estimates for Multi-class Classification by Pairwise Coupling

The Journal of Machine Learning Research
Pairwise classifier combination using belief functions

Pattern Recognition Letters
Decision making under uncertainty using imprecise probabilities

International Journal of Approximate Reasoning
Joint propagation of probability and possibility in risk analysis: Towards a formal framework

International Journal of Approximate Reasoning
Generalized Bradley-Terry Models and Multi-Class Probability Estimates

The Journal of Machine Learning Research
Refined modeling of sensor reliability in the belief function framework using contextual discounting

Information Fusion
Solving multiclass learning problems via error-correcting output codes

Journal of Artificial Intelligence Research

Correcting binary imprecise classifiers: local vs global approach

SUM'12 Proceedings of the 6th international conference on Scalable Uncertainty Management

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper proposes a simple framework to combine binary classifiers whose outputs are imprecise probabilities (or are transformed into some imprecise probabilities, e.g., by using confidence intervals). This combination comes down to solve linear programs describing constraints over events (here, subsets of classes). The number of constraints grows linearly with the number of classifiers, making the proposed framework tractable for problems involving a relatively large number of classes. After detailing the method, we provide some first experimental results illustrating the method interests.