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
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 changed | Where | Measured | Effect |
|---|---|---|---|
| Balance likelihood and classification objectives at γ = 1 | IB-INN · CIFAR10 | calibration error (Geo. mean): 1.26 vs 6.23 | Improved |
| Balance likelihood and classification objectives at γ = 1 | IB-INN · CIFAR10 | classification error: 10.27 vs 6.51 | Hurt |
| Balance likelihood and classification objectives at γ = 1 | IB-INN · CIFAR10 | OoD detection score (Average): 68.76 vs 73.92 | No clear effect |
| Balance likelihood and classification objectives at γ = 0.1 | IB-INN · CIFAR100 | calibration error (Geo. mean): 2.60 vs 5.13 | Improved |
| Balance likelihood and classification objectives at γ = 0.1 | IB-INN · CIFAR100 | classification error: 42.57 vs 29.27 | Hurt |
| Remove the L_Y classification objective (ablation) | IB-INN · CIFAR10 | bits/dim: 5.25 -> 4.80 | Improved |
| Remove the L_Y classification objective (ablation) | IB-INN · CIFAR100 | bits/dim: 4.94 -> 4.82 | Improved |
| Remove the L_X likelihood objective (ablation) | IB-INN · CIFAR10 | classification error: 10.27 -> 8.72 | Improved |
| Remove the L_X likelihood objective (ablation) | IB-INN · CIFAR10 | bits/dim: 5.25 -> 17.27 | Hurt |
| Remove the L_X likelihood objective (ablation) | IB-INN · CIFAR100 | classification error: 42.57 -> 33.78 | Improved |
| Remove the L_X likelihood objective (ablation) | IB-INN · CIFAR100 | bits/dim: 4.94 -> 18.44 | Hurt |
| Increase γ toward the classification-dominated regime | IB-INN · CIFAR10 | classification error | Improved |
Closest papers by meaning
- Flowing Backwards: Improving Normalizing Flows via Reverse Representation Alignment (AAAI 2026 · similarity 0.87)
- Understanding Likelihood of Normalizing Flow and Image Complexity through the Lens of Out-of-Distribution Detection (AAAI 2024 · similarity 0.86)
- Information Bottleneck Analysis of Deep Neural Networks via Lossy Compression (ICLR 2024 · similarity 0.86)
- Generative Cross-Entropy: A Strictly Proper Loss for Data-Efficient Classification (NeurIPS 2026 · similarity 0.85)
- Understanding and Leveraging the Learning Phases of Neural Networks (AAAI 2024 · similarity 0.85)
- Analysis of Learning a Flow-based Generative Model from Limited Sample Complexity (ICLR 2024 · similarity 0.85)
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.
