Multi-Task Preference Learning with Gaussian Processes



Preference learning is concerned with the acquisition of preference models from data – it involves learning from observations that reveal information about the preferences of an individual or a class of individuals, and building models that generalize beyond such training data. In many cases, data for a single individual is scarce, and it pays to build a hierarchical model of preferences, where models for individuals share a common prior. This setting is referred to as multi-task learning. In particular, we here consider the application of multi-task Gaussian processes to the learning of preferences, where each of the models for individual subjects is a nonparametric Gaussian process. We demonstrate the usefulness of our approach on an audiological data set.
About Dr. Tom Heskes
Dr. Tom Heskes is a Professor in Artificial Intelligence, and he leads the Machine Learning Group, at the Institute for Computing and Information Sciences, Radboud University Nijmegen. His research is on probabilistic methods for machine learning and artificial intelligence, with applications in, among others, neuroscience and bioinformatics. He is a Vici laureate and Editor-in-Chief of Neurocomputing.
Contact information:
Dr. Wolf Ketter