Graphical Multi-Task Learning
Daniel R. Sheldon · eCommons (Cornell University) · 2008
We investigate the problem of learning mul-tiple tasks that are related according to a network structure, using the multi-task ker-nel framework proposed in (Evgeniou et al., 2006). Our method combines a graphical task kernel with an arbitrary base kernel. We demonstrate its effectiveness on a real ecolog-ical application that inspired this work. 1.