AI Paper Atlas

Research map / Probabilistic and scientific modeling / Variational and generative sampling

Training Normalizing Flows with the Information Bottleneck for Competitive Generative Classification

Lynton Ardizzone, Radek Mackowiak, Carsten Rother, Ullrich Köthe

NeurIPS 2020 · Probabilistic and scientific modeling › Variational and generative sampling · approach: Normalizing flow training and design · 6 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
Balance likelihood and classification objectives at γ = 1IB-INN · CIFAR10calibration error (Geo. mean): 1.26 vs 6.23Improved
Balance likelihood and classification objectives at γ = 1IB-INN · CIFAR10classification error: 10.27 vs 6.51Hurt
Balance likelihood and classification objectives at γ = 1IB-INN · CIFAR10OoD detection score (Average): 68.76 vs 73.92No clear effect
Balance likelihood and classification objectives at γ = 0.1IB-INN · CIFAR100calibration error (Geo. mean): 2.60 vs 5.13Improved
Balance likelihood and classification objectives at γ = 0.1IB-INN · CIFAR100classification error: 42.57 vs 29.27Hurt
Remove the L_Y classification objective (ablation)IB-INN · CIFAR10bits/dim: 5.25 -> 4.80Improved
Remove the L_Y classification objective (ablation)IB-INN · CIFAR100bits/dim: 4.94 -> 4.82Improved
Remove the L_X likelihood objective (ablation)IB-INN · CIFAR10classification error: 10.27 -> 8.72Improved
Remove the L_X likelihood objective (ablation)IB-INN · CIFAR10bits/dim: 5.25 -> 17.27Hurt
Remove the L_X likelihood objective (ablation)IB-INN · CIFAR100classification error: 42.57 -> 33.78Improved
Remove the L_X likelihood objective (ablation)IB-INN · CIFAR100bits/dim: 4.94 -> 18.44Hurt
Increase γ toward the classification-dominated regimeIB-INN · CIFAR10classification errorImproved

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