Code for our recent paper: Yongxiang Li, Yuting Pu, Changming Cheng, and Qian Xiao. (2023). A Scalable Gaussian Process for Large-Scale Periodic Data. Technometrics. 65(3), 363–374. Link: https://www.tandfonline.com/doi/full/10.1080/00401706.2023.2166124
Abstract:
The periodic Gaussian process (PGP) has been increasingly used to model periodic data due to its high accuracy. Yet, computing the likelihood of PGP has a high computational complexity of
Instructions:
- Please download and install MATLAB to use this code
- Add \routine\ as well as \smt\ package to your workpath
- fit_CPGP.m in the routine package is the main function to fit scalable CPGP model.
- smt package: a Matlab toolbox for structured matrices. NUMERICAL ALGORITHMS, 59 (2012), pp. 639-659 DOI: 10.1007/s11075-011-9527-9. this package is used for Toeplitz matrices' fast computation
Notes:
- The input data for fitting CPGP model is supposed to be grid
- routine package also contains a comparison method NRCPE.m
- The results in simulation and real case study can be reproduced by the code in corresponding folders
- The raw experimental data are also included (Public dataset like CWRU is not included)
Citation: If you find our work useful in your research, please consider citing: @article{li2023scalable, title={A scalable Gaussian process for large-scale periodic data}, author={Li, Yongxiang and Pu, Yuting and Cheng, Changming and Xiao, Qian}, journal={Technometrics}, volume={65}, number={3}, pages={363–374}, year={2023}, publisher={Taylor & Francis} }