TY - GEN
T1 - Trustworthy Prediction with Gaussian Process Knowledge Scores
AU - Butler, Kurt
AU - Feng, Guanchao
AU - Chen, Tong
AU - Djurić, Petar M.
N1 - Publisher Copyright:
© 2025 European Signal Processing Conference, EUSIPCO. All rights reserved.
PY - 2025
Y1 - 2025
N2 - Probabilistic models are often used to make predictions in regions of the data space where no observations are available, but it is not always clear whether such predictions are well-informed by previously seen data. In this paper, we propose a knowledge score for predictions from Gaussian process regression (GPR) models that quantifies the extent to which observing data have reduced our uncertainty about a prediction. The knowledge score is interpretable and naturally bounded between 0 and 1. We demonstrate in several experiments that the knowledge score can anticipate when predictions from a GPR model are accurate, and that this anticipation improves performance in tasks such as anomaly detection, extrapolation, and missing data imputation. Source code for this project is available online at https://github.com/KurtButler/GP-knowledge.
AB - Probabilistic models are often used to make predictions in regions of the data space where no observations are available, but it is not always clear whether such predictions are well-informed by previously seen data. In this paper, we propose a knowledge score for predictions from Gaussian process regression (GPR) models that quantifies the extent to which observing data have reduced our uncertainty about a prediction. The knowledge score is interpretable and naturally bounded between 0 and 1. We demonstrate in several experiments that the knowledge score can anticipate when predictions from a GPR model are accurate, and that this anticipation improves performance in tasks such as anomaly detection, extrapolation, and missing data imputation. Source code for this project is available online at https://github.com/KurtButler/GP-knowledge.
KW - anomaly detection
KW - Gaussian processes
KW - predictive distributions
KW - regression models
KW - trustworthy machine learning
UR - https://www.scopus.com/pages/publications/105029832086
U2 - 10.23919/EUSIPCO63237.2025.11226511
DO - 10.23919/EUSIPCO63237.2025.11226511
M3 - Conference contribution
AN - SCOPUS:105029832086
T3 - European Signal Processing Conference
SP - 875
EP - 879
BT - 2025 33rd European Signal Processing Conference, EUSIPCO 2025 - Proceedings
PB - European Signal Processing Conference, EUSIPCO
T2 - 33rd European Signal Processing Conference, EUSIPCO 2025
Y2 - 8 September 2025 through 12 September 2025
ER -