README.md~

# Multi-linear Gaussian Process
A Gaussian process model that learns high-order structure in the data using multi-linear (tensor) kernel. 

[![Build Status](https://travis-ci.org/voxpelli/node-github-publish.svg?branch=master)](https://travis-ci.org/voxpelli/node-github-publish)
[![Coverage Status](https://coveralls.io/repos/voxpelli/node-github-publish/badge.svg)](https://coveralls.io/r/voxpelli/node-github-publish)
[![Dependency Status](https://gemnasium.com/voxpelli/node-github-publish.svg)](https://gemnasium.com/voxpelli/node-github-publish)

Multilinear Gaussian process regression, implemented in MATLAB.
See details in our paper [Tensor Regression meets Gaussian Process](https://arxiv.org/abs/1711.00073)

![](tlstm.png "Graphical Model for MLGP")

# File  

**test example**

```matlab
run test_mlgp.m
```



## Directory

* **reader.py**
read the data into train/valid/test datasets, normalize the data if needed

* **model.py**
seq2seq model for sequence prediction

* **trnn.py**
tensor-train lstm cell and corresponding tensor train contraction

* **trnn_imply.py**
forward step in tensor-train rnn, feed previous predictions as input

## Citation

If you think the repo is useful, we kindly ask you to cite our work at 

```
@article{yu2017long,
  title={Long-term forecasting using tensor-train RNNs},
  author={Yu, Rose and Zheng, Stephan and Anandkumar, Anima and Yue, Yisong},
  journal={arXiv preprint arXiv:1711.00073},
  year={2017}
}
```