Multitask learning Jaakko Peltonen, 16.10.2007 Summary Often, there is too little data available about a learning task. This means that learning a complicated model is likely to overfit. Learning a simple model might not overfit, but likely won't tell enough about the underlying relationships in the data. However, often there is data available for several related problems. If we can find the relationships between the problems, we can use the relationships to share information, and learn the problems better than by learning them separately. Learning all these problems at the same time is called multitask learning.