Research map / Probabilistic and scientific modeling / Variational and generative sampling
InfoBridge: Mutual Information estimation via Bridge Matching
Sergei Kholkin, Ivan Butakov, Evgeny Burnaev, Nikita Gushchin, Alexander Korotin
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 |
|---|---|---|---|
| use bridge matching (domain transfer) instead of diffusion generative modeling for MI | MINDE-C · Image MI benchmark (Gaussians and Rectangles, 16x16 and 32x32) | MAE: 0.56 -> 0.38 | Improved |
| use bridge matching MI estimator | MINDE-C · Protein embeddings (ProtTrans5) | MAE: 9.29 -> 0.04 | Improved |
| use bridge matching MI estimator | MIENF · Image MI benchmark (Gaussians and Rectangles, 16x16 and 32x32) | MAE: 0.45 -> 0.38 | Improved |
| bridge matching MI estimator vs diffusion MINDE | MINDE-C · 16x16 Rectangles, N=128 | std of MI estimate: 6.65 -> 0.72 | Improved |
| bridge matching MI estimator vs diffusion MINDE | MINDE-C · 16x16 Rectangles, N=131072 | std of MI estimate: 0.75 -> 0.028 | Improved |
| use bridge matching on high-MI tasks | MINE, InfoNCE, fDIME · High MI benchmark (d up to 160, MI up to 80) | MI estimation accuracy | Improved |
| use bridge matching on low-dimensional benchmark | MINDE · Low-dimensional benchmark | MI estimate | No clear effect |
| apply arcsinh tail-shortening transform (ablation) | InfoBridge · Cauchy / Student-t dof=1 low-dim | MI estimate | Improved |
| use bridge matching on heavy-tailed Cauchy data without transform | InfoBridge · Student-t dof=1 | MI estimate | Hurt |
Closest papers by meaning
- A Benchmark Suite for Evaluating Neural Mutual Information Estimators on Unstructured Datasets (NeurIPS 2024 · similarity 0.86)
- InfoNet: Neural Estimation of Mutual Information without Test-Time Optimization (ICML 2024 · similarity 0.86)
- FMMI: Flow Matching Mutual Information Estimation (NeurIPS 2026 · similarity 0.86)
- MINDE: Mutual Information Neural Diffusion Estimation (ICLR 2024 · similarity 0.86)
- Mutual Information Estimation via $f$-Divergence and Data Derangements (NeurIPS 2024 · similarity 0.85)
- Approximating mutual information of high-dimensional variables using learned representations (NeurIPS 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.
