Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Bayesian Pseudocoresets
Dionysis Manousakas, Zuheng Xu, Cecilia Mascolo, Trevor Campbell
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 |
|---|---|---|---|
| replace SparseVI/Hilbert coreset (original datapoints) with learned pseudodata (PSVI) | Bayesian coreset variational inference · Gaussian mean inference, d=200 and d=500 | KL divergence to true posterior | Improved |
| increase data dimension from 200 to 500 | previous coreset methods · Gaussian mean inference | KL divergence | Hurt |
| use PSVI instead of SparseVI | coreset construction · Bayesian linear regression, d=100 | normalized KL divergence | Improved |
| use Hilbert coresets (GIGA) with random projections | coreset construction · Bayesian linear regression | normalized KL divergence | No clear effect |
| use PSVI instead of SparseVI at small coreset size | coreset construction · Bayesian logistic regression on TRANSACTIONS, CHEMREACT100, MUSIC | reverse KL divergence (normalized by prior) | Improved |
| use GIGA (Optimal) instead of PSVI when coreset size M exceeds d | coreset construction · TRANSACTIONS (d=50) | KL divergence | Improved |
| add differential privacy (clipping and noise) to PSVI (ablation) | PSVI · logistic regression, 3 large datasets, (0.2, 1/N)-DP | posterior approximation quality (KL) | No clear effect |
| use DP-PSVI instead of DP-VI in the high-privacy regime | private Bayesian inference · logistic regression on TRANSACTIONS, CHEMREACT100, MUSIC | KL divergence | Improved |
Closest papers by meaning
- Fast Bayesian Coresets via Subsampling and Quasi-Newton Refinement (NeurIPS 2022 · similarity 0.90)
- Automated Scalable Bayesian Inference via Hilbert Coresets (JMLR 2019 · similarity 0.89)
- Coresets for Scalable Bayesian Logistic Regression (NeurIPS 2016 · similarity 0.89)
- General bounds on the quality of Bayesian coresets (NeurIPS 2024 · similarity 0.87)
- On Inherent Privacy of Posterior Sampling: A Unified R\'enyi-Divergence Framework (NeurIPS 2026 · similarity 0.85)
- Beyond Worst-Case Coreset Bounds for $k$-Clustering via Determinantal Sampling (NeurIPS 2026 · similarity 0.84)
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.
