This is the implementation of Near-Optimal Active Learning of Multi-Output Gaussian Processes paper (https://arxiv.org/abs/1511.06891). This repository uses the Intrinsic Coregionalization Model (ICM) instead of Convolved MOGP model used in the paper.
After installing the listed dependencies, simply clone this package to run scripts.
See run.py script to setup the Jura dataset used in the paper and perform active learning with MOGP. You can setup certain arguments from command line. Simply execute the following to start the training.
python run.py
The plot of MAE (Mean absolute error) of model's prediction of Cd concentration on the test set against the number of training samples is shown below:
For any queries, feel free to raise an issue or contact me at [email protected].
This project is licensed under the MIT License - see the LICENSE.md file for details.