This repository is the official implementation of the methods in the publication:

• P.E. Chang, P. Verma, S.T. John, A. Solin, and M.E. Khan (2023). Memory-based dual Gaussian processes for sequential learning. In International Conference on Machine Learning (ICML). [arXiv]

Sequential learning with Gaussian processes (GPs) is challenging when access to past data is limited, for example, in continual and active learning. In such cases, errors can accumulate over time due to inaccuracies in the posterior, hyperparameters, and inducing points, making accurate learning challenging. Here, we present a method to keep all such errors in check using the recently proposed dual sparse variational GP. Our method enables accurate inference for generic likelihoods and improves learning by actively building and updating a memory of past data. We demonstrate its effectiveness in several applications involving Bayesian optimization, active learning, and continual learning.

We recommend setting up a conda environment for running the experiments. The code base is tested on a machine with a Ubuntu 22.04 distribution, CUDA11.6, and conda 23.1.0.

conda create -n sequential-gp python==3.8
conda activate sequential-gp

Within the virtual environment, install the dependencies by running

pip install -r requirements.txt

(Note that the hotspots experiment has its own environment and setup instructions.)

There are a series of experiments which are organized inside the experiments folder as separate sub-folders. Each experiment sub-folder has their respective readme files with instructions on how to run the particular experiment.

The datasets used for banana, UCI, and magnetometer experiments are available in experiments/data/ directory. The original source of the data sets are:

• Banana: https://github.com/thangbui/streaming_sparse_gp
• UCI: https://archive.ics.uci.edu/datasets
• Magnetometer: https://github.com/AaltoML/magnetic-data

For all correspondence, please contact [email protected] or [email protected].

This software is provided under the MIT license.