Chapter Summary
We have now reviewed the main issues encountered when combining predictors in groups.
We first introduced the Supervised Learning problem and described the multilayer perceptron, the predictor to be used in this thesis. We then reviewed some of the methods for
combining a given set of predictors, followed by a summary of architectures for learning the
predictors that are to be combined. Among these architectures is the paradigm of ensemble
learning, where each predictor attempts to solve the same problem and errors can potentially be reduced by averaging their predictions. Many of these techniques will be described
in more detail in a taxonomy we propose in the next chapter.