11.5 Metalearning Algorithms 477 appropriate number of iterations can be determined using internal cross-validation; there is a shrinkage parameter that can be tuned to prevent overfitting; also, you can choose resampling instead of reweighting. RacedIncrementalLogitBoost learns by racing LogitBoosted committees, and operates incrementally by processing the data in batches, making it useful for large datasets (Frank et al., 2002). Each committee member is learned from a different batch. The batch size starts at a given minimum and repeatedly doubles until it reaches a preset maximum. Resampling is used if the base classifier cannot handle weighted instances (you can force resampling anyway). Log-likelihood pruning can be used within each committee: This discards new com- mittee members if they decrease the log-likelihood based on the validation data. It is up to you to determine how many instances to hold out for validation. The valida- tion data is also used to determine which committee to retain when training terminates. CombiningClassifiers Vote provides a baseline method for combining classifiers. The default scheme is to average their probability estimates or numeric predictions, for classification and