Real-World Applications

Subseasonal climate forecasting develops a probabilistic framework for Madden-Julian Oscillation (MJO) prediction based on GP models with empirical correlations and a covariance correction, extending probabilistic MJO coverage to over three weeks. Additionally, the method avoids the need for hyperparameter optimization, streamlining the process and enhancing efficiency and stability

Privacy-aware regression develops a comprehensive theoretical and methodological framework for privacy-aware Gaussian process regression that addresses the fundamental privacy-utility trade-off inherent in predictive modeling with confidential data. The proposed method enforces variance-based privacy guarantees by solving for an optimal synthetic noise covariance via semi-definite programming, so GP predictors stay accurate yet satisfy the prescribed privacy level in settings such as satellite tracking and census analytics.