Linear Combiners for Classifier Fusion: Some Theoretical and Experimental Results

In this paper, we continue the theoretical and experimental analysis of two widely used combining rules, namely, the simple and weighted average of classifier outputs, that we started in previous works. We analyse and compare the conditions which affect the performance improvement achievable by weighted average over simple average, and over individual classifiers, under the assumption of unbiased and uncorrelated estimation errors. Although our theoretical results have been obtained under strict assumptions, the reported experiments show that they can be useful in real applications, for designing multiple classifier systems based on linear combiners.