Distributed Learning for Grid Spectral Mixture (GSM) kernel

This repository contains the code and data used in the paper "Gaussian Process Regression with Grid Spectral Mixture Kernel: Distributed Learning for Multidimensional Data" published at the 25th International Conference on Information Fusion (FUSION) 2022.

• R. C. Suwandi, Z. Lin, Y. Sun, Z. Wang, L. Cheng and F. Yin, "Gaussian Process Regression with Grid Spectral Mixture Kernel: Distributed Learning for Multidimensional Data," 2022 25th International Conference on Information Fusion (FUSION), 2022, pp. 1-8, doi: 10.23919/FUSION49751.2022.9841347.

Citation

@INPROCEEDINGS{9841347,  
  author={Suwandi, Richard Cornelius and Lin, Zhidi and Sun, Yiyong and Wang, Zhiguo and Cheng, Lei and Yin, Feng},  
  booktitle={2022 25th International Conference on Information Fusion (FUSION)},   
  title={Gaussian Process Regression with Grid Spectral Mixture Kernel: Distributed Learning for Multidimensional Data},   
  year={2022},  
  volume={},  
  number={},  
  pages={1-8},  
  doi={10.23919/FUSION49751.2022.9841347}
}

Example

To see an example, please run the dsca.m/dsca_multi.m file for the Distributed SCA (DSCA) algorithms and the d2sca.m/d2sca_multi.m file for the Doubly Distributed SCA (D2SCA) algorithms. To change the data set, simply uncomment one of the listed data sets and comment the others, e.g., for the Electricity data set,

% Read in data & some general setup
file_name = 'electricitydata';
% file_name = 'passengerdata';
% file_name = 'hoteldata';
% file_name = 'employmentdata';
% file_name = 'unemployment';
% file_name = 'clay';
% file_name = 'CO2';
% file_name = 'ECG_signal';  

disp(['Simulation on ',file_name]);
[xtrain, ytrain, xtest, ytest] = load_data(file_name);
nTrain = length(xtrain);
nTest = 20;

Other algorithm setups can also be changed in the corresponding files.

Dependencies

The current version of the code uses MATLAB R2021a (https://www.mathworks.com/products/matlab.html) and MOSEK version 9.3 (https://docs.mosek.com/9.3/install/installation.html). Please refer to the corresponding websites for the installation instructions.