Research map / Probabilistic and scientific modeling / Variational and generative sampling
Variational Inference and Model Selection with Generalized Evidence Bounds
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Replace ELBO-based importance-weighted objective with CLBO (chi-evidence bound, T=200) | VAE · MNIST / OMNIGLOT / Frey / Caltech test log-likelihood | test log-likelihood: e.g. MNIST L=1 K=5: -84.71 vs IW-VAE -85.41 | Helped |
| Use CLBO (T=2) instead of ELBO/Rényi objective for Bayesian NN regression | Bayesian neural net regression · 10 UCI datasets | test RMSE and log-likelihood: Boston RMSE 2.71 vs Rényi 2.86 | Helped |
| Apply GLBO objective to AVB | AVB on MNIST · MNIST | ELBO: -79.97 vs -82.7 | Helped |
| Increase K (number of samples) or decrease temperature T in CLBO (ablation) | CLBO on toy distribution · toy bound sharpness | bound value | Helped |
| Use Rényi VI bound (RVB) with T>1 instead of importance-weighted ELBO (ablation) | RVB on toy distribution · toy bound sharpness | bound value | Hurt |
| Model-selection refinement phase with saturating log-log evidence function after AVB convergence (ablation) | AVB on MNIST · MNIST train/test log-evidence | log-evidence: testing +1.47 vs training +0.58 nats | Helped |
| Use high SNR update rule (separate K-sample objectives for alpha and beta) on vanilla ELBO-VAE (ablation) | VAE on CelebA · CelebA | log-evidence | No clear effect |
Closest papers by meaning
- Importance Weighted Hierarchical Variational Inference (NeurIPS 2019 · similarity 0.91)
- Sticking the Landing: Simple, Lower-Variance Gradient Estimators for Variational Inference (NeurIPS 2017 · similarity 0.91)
- Predictive variational inference: Learn the predictively optimal posterior distribution (ICML 2026 · similarity 0.91)
- Variational Inference with Locally Enhanced Bounds for Hierarchical Models (ICML 2022 · similarity 0.91)
- SoftCVI: Contrastive variational inference with self-generated soft labels (ICLR 2025 · similarity 0.89)
- Maxitive Donsker-Varadhan Formulation for Possibilistic Variational Inference (NeurIPS 2026 · similarity 0.89)
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.