Guest Introduction: Special Issue on New Methods for Model Selection and Model Combination

A classical challenge in fitting models to data is managing the complexity of the models to simultaneously avoid under-fitting and over-fitting the data; fulfilling the goal of producing models that generalize well to unseen data. The papers in this special issue present recent developments in model-complexity control for supervised learning. These thirteen papers represent three areas of significant current research on this subject: model selection (explicitly choosing model complexity), sparse models (reducing complexity by enforcing sparse representations), and model combination (combining multiple models to improve generalization).