AI Paper Atlas

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

Efficient Bayesian Inference of Sigmoidal Gaussian Cox Processes

Christian Donner, Manfred Opper

JMLR 2018 · Probabilistic and scientific modeling › Uncertainty and Gaussian processes · approach: Sparse variational Gaussian processes · 18 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
use sigmoid link instead of squared link (mean-field vs Lloyd et al.)Gaussian Cox process variational inference · 1D synthetic, N=453log expected test likelihood: 710.43 -> 720.81Improved
use sigmoid link instead of squared linkGaussian Cox process variational inference · 1D synthetic, N=47log expected test likelihood: -44.26 -> -43.43No clear effect
use sigmoid link instead of squared linkGaussian Cox process variational inference · 1D synthetic, N=4652log expected test likelihood: 17496.75 -> 17497.31No clear effect
use sigmoid link instead of squared linkGaussian Cox process variational inference · 1D synthetic, N=453RMSE: 2.11 -> 0.97Improved
use sigmoid mean-field instead of squared linkGaussian Cox process variational inference · 1D synthetic, N=4652runtime (s): 0.79 -> 0.41Improved
use sigmoid mean-field instead of squared linkGaussian Cox process variational inference · 1D synthetic, N=453runtime (s): 0.23 -> 0.50Hurt
use mean-field or Laplace instead of variational Gauss for sigmoidal Cox processSigmoidal Gaussian Cox process · 1D generative-model exampleruntime (s): 80 -> 0.09Improved
use approximate inference instead of MCMC samplerSigmoidal Gaussian Cox process · 1D generative-model data (10 datasets)log expected test likelihood: 875.5 -> 686.2Hurt
increase number of inducing points beyond about 49 (ablation)Mean field, Laplace, variational Gauss · 1D syntheticlog expected test likelihoodNo clear effect
increase number of integration points (ablation)Mean field, Laplace, variational Gauss · 2D synthetic (10x10 inducing)log expected test likelihoodNo clear effect
warm-start by first fitting with 1000 integration points then refitting with target number (ablation)Sigmoidal mean-field · Mouse neuron and Porto taxi datasetsruntime / test likelihoodImproved
use sigmoid link instead of squared linkGaussian Cox process variational inference · Mouse neuron 2D datalog expected test likelihoodHurt

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