MTGP - A Multi-task Gaussian Process Toolbox

Authors

Robert Dürichen (University of Lübeck)
Marco AF Pimentel (University of Oxford)
Lei Clifton (University of Oxford)
Achim Schweikard (University of Lübeck)
David A Clifton (University of Oxford)

Abstract

Gaussian process (GP) models are a flexible means of performing non-parametric Bayesian regression. However, the majority of existing work using GP models in healthcare data is defined for univariate output time-series, denoted as single-task GPs (STGP). Here, we investigate how GPs could be used to model multiple correlated univariate physiological time-series simultaneously. The resulting multi-task GP (MTGP) framework can learn the correlation within multiple signals even though they might be sampled at different frequencies and have training sets available for different intervals. We illustrate the basic properties of MTGPs using a synthetic case-study with respiratory motion data. Finally, two real-world biomedical problems are investigated from the field of patient monitoring and motion compensation in radiotherapy. The results are compared to STGPs and other standard methods in the respective fields. In both cases, MTGPs learned the correlation between physiological time-series efficiently, which leads to improved modelling accuracy.

Download

The current release is v1.4: download "MTGP" toolbox for Matlab

As well as downloading the MTGP toolbox, you will need:
[GPML] - v3.4 or above; a Matlab toolbox for Gaussian processes.

Using this toolbox should be straightforward: the download comes with some toy datasets on which the demos, shown below can be run. Alternatively, use the scripts in the "example" folder to perform the same. Please feel free to contact the authors for more details concerning any of the scripts, or if you encounter any difficulties in using the toolbox.

Copyright

The documents contained in these directories are included by the contributing authors as a means to ensure timely dissemination of scholarly and technical work on a non-commercial basis. This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with this program. If not, see http://www.gnu.org/licenses/.

Literature

Dürichen, R., Pimentel, M.A.F., Clifton, L., Schweikard, A., and Clifton, D.A.:
Multi-task Gaussian Process Models for Biomedical Applications
IEEE Biomedical & Health Informatics, Valencia, Spain, 2014, pp. 492-495 [PDF]

Dürichen, R., Wissel, T., Ernst, F., Pimentel, M.A.F., Clifton, D.A., and Schweikard, A.:
Unified Approach for Respiratory Motion Prediction and Correlation with Multi-task Gaussian Processes
IEEE MLSP, Reims, France, 2014, pp. 1-6 [PDF]

Dürichen, R., Pimentel, M.A.F., Clifton, L., Schweikard, A., and Clifton, D.A.: Multi-task Gaussian Processes for Multivariate Physiological Time-Series Analysis IEEE Transactions on Biomedical Engineering 62(1), 2015, pp. 314 - 322 [PDF]

Verantwortlich