Manifold Gaussian Processes for Regression
This addresses the limitation of smoothness assumptions in GPs for practitioners in fields like robotics, though it is an incremental improvement over existing feature space methods.
The paper tackles the problem of modeling complex, non-differentiable functions with Gaussian Processes by proposing Manifold Gaussian Processes, which jointly learns a data transformation and regression, resulting in improved performance on step functions and robotics tasks where standard GPs fail.
Off-the-shelf Gaussian Process (GP) covariance functions encode smoothness assumptions on the structure of the function to be modeled. To model complex and non-differentiable functions, these smoothness assumptions are often too restrictive. One way to alleviate this limitation is to find a different representation of the data by introducing a feature space. This feature space is often learned in an unsupervised way, which might lead to data representations that are not useful for the overall regression task. In this paper, we propose Manifold Gaussian Processes, a novel supervised method that jointly learns a transformation of the data into a feature space and a GP regression from the feature space to observed space. The Manifold GP is a full GP and allows to learn data representations, which are useful for the overall regression task. As a proof-of-concept, we evaluate our approach on complex non-smooth functions where standard GPs perform poorly, such as step functions and robotics tasks with contacts.