Research map / Optimization / Optimizers
Markov Chain Score Ascent: A Unifying Framework of Variational Inference with Markovian Gradients
Kyurae Kim, Jisu Oh, Jacob R. Gardner, Adji Bousso Dieng, Hongseok Kim
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 pMCSA instead of ELBO N=10 | mean-field BNN · BNN regression, UCI energy | test log predictive density: -2.40 -> -1.92 | Improved |
| use pMCSA instead of ELBO N=10 | mean-field BNN · BNN regression, UCI airfoil | test log predictive density: -2.56 -> -2.27 | Improved |
| use pMCSA instead of ELBO N=10 | mean-field BNN · BNN regression, UCI gas | test log predictive density: 0.13 -> -0.09 | Hurt |
| use pMCSA instead of ELBO N=10 | mean-field BNN · BNN regression, UCI yacht | test log predictive density: -2.44 -> -2.49 | No clear effect |
| use pMCSA instead of unlabeled MCSA variant 2 | mean-field BNN · BNN regression, UCI sml | test log predictive density: -1.97 -> -1.22 | Improved |
| use pMCSA instead of unlabeled MCSA variant 1 | mean-field BNN · BNN regression, UCI gas | test log predictive density: -0.47 -> -0.09 | Improved |
| use pMCSA instead of unlabeled MCSA variant 1 | mean-field BNN · BNN regression, UCI energy | test log predictive density: -2.38 -> -1.92 | Improved |
| use pMCSA instead of unlabeled MCSA variant 1 | mean-field BNN · BNN regression, UCI wine | test log predictive density: -0.97 -> -0.95 | No clear effect |
| increase ELBO budget from N=1 to N=10 | ELBO · BNN regression, UCI energy | test log predictive density: -2.42 -> -2.40 | No clear effect |
| increase ELBO budget from N=1 to N=10 | ELBO · BNN regression, UCI gas | test log predictive density: -0.06 -> 0.13 | Improved |
| increase computational budget N (2^3 to 2^7) for pMCSA | pMCSA · 50-D Gaussian, gradient variance | gradient variance | Improved |
| increase computational budget N for MSC | MSC · 50-D Gaussian, gradient variance | gradient variance | No clear effect |
Closest papers by meaning
- Markovian Score Climbing: Variational Inference with KL(p||q) (NeurIPS 2020 · similarity 0.91)
- On Markov Chain Gradient Descent (NeurIPS 2018 · similarity 0.89)
- Structured Stochastic Gradient MCMC (ICML 2022 · similarity 0.89)
- Stochastic Gradient Descent as Approximate Bayesian Inference (JMLR 2017 · similarity 0.89)
- Stochastic Gradient Descent under Markovian Sampling Schemes (ICML 2023 · similarity 0.89)
- A Variational Analysis of Stochastic Gradient Algorithms (ICML 2016 · similarity 0.88)
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.
