Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Finite-Dimensional BFRY Priors and Variational Bayesian Inference for Power Law Models
Juho Lee, Lancelot F James, Seungjin Choi
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 FNSPM (stable prior) instead of FDPM with collapsed Gibbs (ablation) | CG · PYP synthetic | test log-likelihood: -33.2078 -> -33.1039 | Improved |
| use FNSPM instead of FDPM with VB (ablation) | VB · PYP synthetic | test log-likelihood: -33.4480 -> -33.1861 | Improved |
| use FNSPM instead of FDPM with CVB (ablation) | CVB · PYP synthetic | test log-likelihood: -33.4278 -> -33.2031 | Improved |
| use FNSPM instead of FDPM with collapsed Gibbs (ablation) | CG · AP corpus | test log-likelihood: -157.2228 -> -157.1920 | Improved |
| use FNSPM instead of FDPM with VB (ablation) | VB · NIPS corpus | test log-likelihood: -352.9104 -> -352.6078 | Improved |
| use FNSPM instead of FDPM with CG (ablation) | CG · CRP synthetic | test log-likelihood: -25.4076 -> -25.4079 | No clear effect |
| use FNSPM instead of FDPM with VB (ablation) | VB · CRP synthetic | test log-likelihood: -25.4148 -> -25.5076 | Hurt |
| use variational (VB/S) instead of collapsed Gibbs (CG/S) (ablation) | FNSPM · PYP synthetic | test log-likelihood: -33.1039 -> -33.1861 | Hurt |
| estimate alpha with VB/S instead of CG/S (ablation) | FNSPM · PYP synthetic | alpha estimate (true 0.7): 0.6940 -> 0.4640 | Hurt |
| estimate alpha with CVB/S (ablation) | FNSPM · PYP synthetic | alpha estimate (true 0.7): 0.7041 | Improved |
| increase finite dimension K from 200 to 1000 (ablation) | finite-dimensional FNSPM algorithms · PYP synthetic | test log-likelihood gap to infinite-dimensional CG | Improved |
| use variational instead of sampling algorithms (ablation) | finite FNSPM · PYP synthetic | running time per iteration | Improved |
Closest papers by meaning
- Sparse Variational Student-t Processes (AAAI 2024 · similarity 0.88)
- Beyond the Chinese Restaurant and Pitman-Yor processes: Statistical Models with double power-law behavior (ICML 2019 · similarity 0.87)
- Scalable and Adaptive Variational Bayes Methods for Hawkes Processes (JMLR 2025 · similarity 0.87)
- Variational Inference for Latent Variables and Uncertain Inputs in Gaussian Processes (JMLR 2016 · similarity 0.87)
- Posterior Asymptotics for Boosted Hierarchical Dirichlet Process Mixtures (JMLR 2022 · similarity 0.87)
- Generic Inference in Latent Gaussian Process Models (JMLR 2019 · similarity 0.87)
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.
