High-order correlations are ubiquitous in modern data analytics. MLGP is a Gaussian process model that learns high-order structure in the data using multi-linear (tensor). kernel.
Multilinear Gaussian process regression, implemented in MATLAB. See details in our AISTATS 2018 paper Tensor Regression meets Gaussian Processes
example dataset
The Restaurant & Consumer Dataset contains data to build a restaurant recommender system where the objective is to predict consumer ratings given to different restaurants. Each of the p3 = 138 consumers gave p2 = 3 scores for food quality, service quality and overall quality. The dataset also contains p1 = 45 various descriptive attributes of the restaurants (such as geographical position, cuisine type and price band). We consider this to be a regression problem where the objective is to predict the scores given the attributes of a restaurant as an input query. Since there are 138 consumers, this leads to a multitask problem composed of 138x3 regression tasks
testing script
run test_mlgp.mIf you think the repo is useful, we kindly ask you to cite our work at
@article{yu2017tensor,
title={Tensor Regression Meets Gaussian Processes},
author={Yu, Rose and Li, Guangyu and Liu, Yan},
booktitle={Artificial Intelligence and Statistics},
year={2018}
}