Learning Coordination Classifiers

We present a new approach to ensemble classifica- tion that requires learning only a single base clas- sifier. The idea is to learn a classifier that simulta- neously predicts pairs of test labels—as opposed to learning multiple predictors for single test labels— then coordinating the assignment of individual la- bels by propagating beliefs on a graph over the data. We argue that the approach is statistically well mo- tivated, even for independent identically distributed (iid) data. In fact, we present experimental results that show improvements in classification accuracy over single-example classifiers, across a range of iid data sets and over a set of base classifiers. Like boosting, the technique increases representational capacity while controlling variance through a prin- cipled form of classifier combination.