A base learner is an individual model trained as one member of an ensemble, whose predictions are combined with those of other base learners to produce a final output. Base learners are typically simple or weak relative to the ensemble as a whole - for example shallow decision trees in a random forest or boosting sequence - so that their errors are diverse and can partially cancel when aggregated. The choice of base learner architecture and training procedure directly affects the diversity and accuracy of the resulting ensemble.