Classification by pairwise coupling of imprecise probabilities

作者：

Highlights：

• We address combining imprecise classifiers in the framework of imprecise probabilities.

• Classifier outputs are probability intervals, which are large for irrelevant classifiers.

• The combination consists in computing probabilities consistent with the intervals.

• This approach is particularly interesting when combining few classifiers.

摘要

•We address combining imprecise classifiers in the framework of imprecise probabilities.•Classifier outputs are probability intervals, which are large for irrelevant classifiers.•The combination consists in computing probabilities consistent with the intervals.•This approach is particularly interesting when combining few classifiers.

论文关键词：Classifier combination,Reasoning under uncertainty,Cautious predictions

论文评审过程：Received 24 November 2016, Revised 6 September 2017, Accepted 15 October 2017, Available online 18 October 2017, Version of Record 6 February 2018.

论文官网地址：https://doi.org/10.1016/j.patcog.2017.10.019