A supervised fuzzy measure learning algorithm for combining classifiers

Fuzzy measure-based aggregations allow taking interactions among coalitions of the input sources into account. Their main drawback when applying them in real-world problems, such as combining classifier ensembles, is how to define the fuzzy measure that governs the aggregation and specifies the interactions. However, their usage for combining classi-fiers has shown its advantage. The learning of the fuzzy measure can be done either in a supervised or unsupervised manner. This paper focuses on supervised approaches. Existing supervised approaches are designed to minimize the mean squared error cost function, even for classification problems. We propose a new fuzzy measure learning algo-rithm for combining classifiers that can optimize any cost function. To do so, advancements from deep learning frameworks are considered such as automatic gradient computation. Therefore, a gradient-based method is presented together with three new update policies that are required to preserve the monotonicity constraints of the fuzzy measures. The use-fulness of the proposal and the optimization of cross-entropy cost are shown in an exten-sive experimental study with 58 datasets corresponding to both binary and multi-class classification problems. In this framework, the proposed method is compared with other state-of-the-art methods for fuzzy measure learning.(c) 2022 Elsevier Inc. All rights reserved.

Automated summary

Fuzzy measure-based aggregations consider interactions among input source coalitions, but defining the fuzzy measure is a challenge. This paper proposes a new algorithm for learning fuzzy measure that can optimize any cost function, using advancements from deep learning frameworks. Experimental study with 58 datasets shows the effectiveness of the proposed method in optimizing cross-entropy cost for binary and multi-class classification problems, compared to other state-of-the-art methods for fuzzy measure learning.