AI Paper Atlas

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

Dichotomize and Generalize: PAC-Bayesian Binary Activated Deep Neural Networks

Gaël Letarte, Pascal Germain, Benjamin Guedj, Francois Laviolette

NeurIPS 2019 · Probabilistic and scientific modeling › Uncertainty and Gaussian processes · approach: Approximate inference in Bayesian neural networks · 4 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
Add pretraining on half the training data and use the learned weights as the prior (ablation)PBGNet · MNIST 4 vs 9test error: 0.040 -> 0.017Improved
Add pretraining on half the training data and use the learned weights as the prior (ablation)PBGNet · MNIST 5 vs 6test error: 0.025 -> 0.009Improved
Add pretraining on half the training data and use the learned weights as the prior (ablation)PBGNet · Adstest error: 0.168 -> 0.033Improved
Add pretraining on half the training data and use the learned weights as the prior (ablation)PBGNet · Adulttest error: 0.163 -> 0.154Improved
Add pretraining on half the training data and use the learned weights as the prior (ablation)PBGNet · MNIST low vs hightest error: 0.052 -> 0.027Improved
Select hyperparameters using the bound while training on empirical loss (ablation)PBGNet_ℓ · MNIST 4 vs 9test error: 0.018 -> 0.035Hurt
Optimize the PAC-Bayes bound directly rather than empirical loss (ablation)PBGNet · Adstest error: 0.030 -> 0.168Hurt
Optimize the PAC-Bayes bound directly rather than empirical loss (ablation)PBGNet · Adulttest error: 0.149 -> 0.163Hurt
Optimize the PAC-Bayes bound directly rather than empirical loss (ablation)PBGNet · MNIST 4 vs 9test error: 0.018 -> 0.040Hurt
Reduce Monte Carlo sample size during training (ablation)PBGNet · some learning problemstraining performanceMixed

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