Gauss-Legendre Features for Scalable Gaussian Process Regression

Explore a technical lecture on scaling up Gaussian process regression through a quadrature-based approach using low-rank approximation of kernel matrices. Learn how Gauss-Legendre features can generate high-quality kernel approximations using poly-logarithmic features relative to training points, offering significant advantages over random Fourier features methods. Discover the implementation details for effective hyperparameter learning, training, and prediction, particularly beneficial for low-dimensional datasets. Presented by Paz Fink, a Ph.D. candidate in applied mathematics at TAU specializing in numerical linear algebra with kernel methods applications, this talk demonstrates practical numerical experiments showcasing the method's utility in machine learning applications.