Demonstration code of decentralized methods for Gaussian processes [1]. For further details check the paper here. A 3-minute presentation of the paper can be found here.
The code implements:
- Decentralized Nested Pointwise Aggregation of Experts (DEC-NPAE)
- Distributed Nested Pointwise Aggregation of Experts (DIST-NPAE)
The source code of the factorized training and the centralized NPAE [2] can be found in the GRBCM [3] GitHub repository.
Install the gpml toolbox.
Execute:
demo_2D.m
[1] G. P. Kontoudis and D. J. Stilwell, “Decentralized Nested Gaussian Processes for Multi-Robot Systems,” in IEEE International Conference on Robotics and Automation (ICRA), 2021.
[2] D. Rullière, N. Durrande, F. Bachoc, and C. Chevalier, “Nested Kriging predictions for datasets with a large number of observations,” Statistics and Computing, 2018.
[3] H. Liu, J. Cai, Y. Wang, and Y. S. Ong, “Generalized robust Bayesian committee machine for large-scale Gaussian process regression,” in International Conference on Machine Learning (ICML), 2018.
Please open a GitHub issue if you encounter any problem or send me an email at gpkont@vt.edu.