When a series of problems are related, representations derived from learning earlier tasks may be useful in solving later problems. In this paper we propose a novel approach to transfer learning with low-dimensional, non-linear latent spaces. We show how such representations can be jointly learned across multiple tasks in a Gaussian Process framework. When transferred to new tasks with relatively few training examples, learning can be faster and/or more accurate. Experiments on digit recognition and newsgroup classification tasks show significantly improved performance when compared to baseline performance with a representation derived from a semi-supervised learning approach or with a discriminative approach that uses only the target data.
