GPatt: Fast Multidimensional Pattern Extrapolation with Gaussian Processes
This work addresses the problem of scalable pattern extrapolation for researchers and practitioners in machine learning, offering a novel framework that improves upon existing methods, though it builds incrementally on prior Gaussian process techniques.
The authors tackled the challenge of using Gaussian processes for automatic pattern extrapolation on large multidimensional datasets, introducing GPatt, a framework that unifies expressive kernels and fast exact inference. They demonstrated that GPatt outperforms other scalable Gaussian process methods in speed and accuracy, handling up to 383,400 training points without human intervention.
Gaussian processes are typically used for smoothing and interpolation on small datasets. We introduce a new Bayesian nonparametric framework -- GPatt -- enabling automatic pattern extrapolation with Gaussian processes on large multidimensional datasets. GPatt unifies and extends highly expressive kernels and fast exact inference techniques. Without human intervention -- no hand crafting of kernel features, and no sophisticated initialisation procedures -- we show that GPatt can solve large scale pattern extrapolation, inpainting, and kernel discovery problems, including a problem with 383400 training points. We find that GPatt significantly outperforms popular alternative scalable Gaussian process methods in speed and accuracy. Moreover, we discover profound differences between each of these methods, suggesting expressive kernels, nonparametric representations, and exact inference are useful for modelling large scale multidimensional patterns.