Laplace approximations are popular techniques for endowing deep networks with epistemic uncertainty estimates as they can be applied without altering the ...

Oct 31, 2024 · Our method provides improved results where prior knowledge is abundant (as is the case in many scientific inference tasks). At the same time, it ...

We now present FSP-LAPLACE, a method for computing Laplace approximations under interpretable GP priors in function space. Section 3.1 introduces an objective.

Jul 18, 2024 · Laplace approximations are popular techniques for endowing deep networks with epistemic uncertainty estimates as they can be applied without altering the ...

Jul 18, 2024 · The paper proposes the FSP-LAPLACE method for incorporating interpretable Gaussian Process (GP) priors into Bayesian deep learning using the ...

The paper proposes a new approach called "FSP-Laplace" for Bayesian deep learning, which uses function-space priors to improve the Laplace approximation.

FSP-Laplace is a novel method that enhances Bayesian deep learning by incorporating interpretable Gaussian process priors directly in function space, improving ...

FSP-Laplace: Function-Space Priors for the Laplace Approximation in Bayesian Deep Learning. Tristan Cinquin, Marvin Pförtner, Vincent Fortuin, Philipp Hennig ...

FSP-Laplace: Function-Space Priors for the Laplace Approximation in Bayesian Deep Learning. T Cinquin, M Pförtner, V Fortuin, P Hennig, R Bamler. arXiv ...

FSP-Laplace: Function-Space Priors for the Laplace Approximation in Bayesian Deep Learning · no code implementations • 18 Jul 2024 • Tristan Cinquin, Marvin ...