AI Paper Atlas

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

Healing Products of Gaussian Process Experts

Samuel Cohen, Rendani Mbuvha, Tshilidzi Marwala, Marc Deisenroth

ICML 2020 · Probabilistic and scientific modeling › Uncertainty and Gaussian processes · approach: PAC-Bayes and decoupled GP learning · 7 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
Replace uniform expert weights with normalized softmax weights based on predictive variancegPoE/rBCM · Regression benchmarks, clustering partitioning (average)NLPD: 0.923 -> 0.354Improved
Combine experts using a Wasserstein barycenter with softmax-variance weightsBAR_var · Regression benchmarks, clustering partitioning (average)NLPD: 0.923 -> 0.353Improved
Use softmax-variance rather than differential-entropy weightsgPoE/rBCM · MNIST PCA featuresNLPD: 0.878 -> 0.309Improved
Use softmax-variance rather than uniform weightsgPoE · MNIST PCA featuresNLPD: 0.382 -> 0.309Improved
Use temperature T=1 rather than T=100 for variance-softmax weights (ablation)BAR_var · AirfoilNLPD: 0.411 -> 1.219Hurt
Use temperature T=1 rather than T=100 for variance-softmax weights (ablation)gPoE_var · AirfoilNLPD: 0.411 -> 0.510Hurt
Increase the temperature from T=1 to T=100 (ablation)rBCM_var with unnormalized weights · AirfoilNLPD: 0.412 -> 0.833Hurt
Increase points per expert from 100 to 500 (ablation)BAR_var · Kin40kNLPD: -0.365 -> -0.720Improved
Increase points per expert from 100 to 500 (ablation)gPoE_var · Kin40kNLPD: -0.359 -> -0.765Improved
Increase points per expert from 100 to 500 (ablation)rBCM_var · Kin40kNLPD: 0.124 -> -0.679Improved
Use normalized rather than unnormalized softmax weightsrBCM · Regression benchmarks across temperaturesNLPDImproved
Aggregate predictions in f-space rather than y-spaceProduct-of-experts models · Regression benchmarksNLPDImproved

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