Classiier Combining through Trimmed Means and Order Statistics

| Combining the outputs of multiple neural networks has led to substantial improvements in several dii-cult pattern recognition problems. In this article, we introduce and investigate robust combiners, a family of classiiers based on order statistics. We focus our study to the analysis of the decision boundaries, and how these boundaries are aaected by order statistics combiners. In particular, we show that using the ith order statistic, or a linear combination of the ordered classiier outputs is quite beneecial in the presence of outliers or uneven classiier performance. Experimental results on several public domain data sets corroborate these ndings.