AI Paper Atlas

Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes

Variational Gaussian processes for linear inverse problems

Thibault Christophe RANDRIANARISOA, Botond Szabo

NeurIPS 2023 · Probabilistic and scientific modeling › Uncertainty and Gaussian processes · approach: PAC-Bayes and decoupled GP learning · 1 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
halve inducing variables from 6 to 3 (ablation)variational GP posterior · Heat equation, n=8000credible set width: 6 -> 3Hurt
reduce inducing variables from 6 to 3 (ablation)variational GP posterior · Heat equation, n=4000, 50 runsMISE: 6 -> 3Hurt
increase inducing variables from 6 to 12 (ablation)variational GP posterior · Heat equation, n=4000MISE: 6 -> 12No clear effect
increase inducing variables from 6 to 12 (ablation)variational GP posterior · Heat equation, n=4000computation time: 6 -> 12Hurt
halve inducing variables from 10 to 5 (ablation)variational GP posterior · Volterra operator, n=15000posterior mean and credible bands: 10 -> 5Hurt
increase inducing variables from 8 to 16 (ablation)variational GP posterior · Volterra operator, n=4000, 30 runsMISE: 8 -> 16No clear effect
reduce inducing variables from 8 to 4 (ablation)variational GP posterior · Volterra operator, n=4000MISE: 8 -> 4Hurt
use a quarter of optimal inducing variables ceil(m/4) (ablation)variational GP posterior · Radon transform, n=500 and n=5000posterior mean smoothness and band width: 10 -> ceil(10/4); 24 -> ceil(24/4)Hurt
use optimal m from theory instead of exact posterior (ablation)variational GP posterior · Heat equation, n=8000computation time: over half an hour -> 50.5 msImproved
use more inducing variables than theoretical threshold (ablation)variational GP posterior · Volterra phase transition, n 100-10000, m 1-17MISE ratioNo clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy