Gaussian Process Latent Factor Regression for Low-Data, High-Dimensional Output Problems
Researchers have introduced Gaussian Process Latent Factor Regression (GPLFR), a new model designed to predict high-dimensional outputs from limited training data. This method addresses limitations in existing multi-output Gaussian processes and compress-then-predict pipelines by integrating compression and prediction into a single objective. GPLFR has been successfully applied to create the first spatially resolved emulator for global climate models of rocky exoplanets. AI
IMPACT Introduces a novel method for handling complex prediction tasks with limited data, potentially advancing scientific modeling.