Research map / Probabilistic and scientific modeling / Variational and generative sampling
Statistical Inference with Unnormalized Discrete Models and Localized Homogeneous Divergences
Takashi Takenouchi, Takafumi Kanamori
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 empirically localized PS-divergence estimator instead of pair scoring rule (kappa=1) | Poisson model · Poisson, theta0=log(2) and log(10) | n * MSE of parameter | Improved |
| use localized PS-divergence estimator instead of MLE on Boltzmann machine with theta*~N(0, I/d) | Fully visible Boltzmann machine · FVBM d=10, theta*~N(0,I/d) | RMSE and computation time | Improved |
| use localized PS-divergence estimator instead of MLE when true parameter theta*~N(0, I) | Fully visible Boltzmann machine · FVBM d=10, theta*~N(0,I) | RMSE | Hurt |
| add l2 regularization (lambda = 10^-2, 10^-4) to the proposed estimator (ablation) | Localized PS-divergence estimator · FVBM d=10, theta*~N(0,I) | RMSE and computation time: lambda = 0, 10^-2, 10^-4 compared | Improved |
| add l2 regularization to the proposed estimator on hidden-variable Boltzmann machine (ablation) | Localized PS-divergence estimator · Boltzmann machine with 2 hidden units, d1=10 | averaged log-likelihood (train/test) | Improved |
| use localized PS-divergence estimator instead of MLE on hidden-variable Boltzmann machine | Boltzmann machine with hidden variables · d1=10, d2=2 | computation time | Improved |
| increase dimension d | Localized PS-divergence estimator · FVBM sample complexity | examples for fixed KL error: approximately 2 ~ 3 times as many examples | Hurt |
Closest papers by meaning
- Estimators for Multivariate Information Measures in General Probability Spaces (NeurIPS 2018 · similarity 0.89)
- Smooth $p$-Wasserstein Distance: Structure, Empirical Approximation, and Statistical Applications (ICML 2021 · similarity 0.87)
- Practical and Consistent Estimation of f-Divergences (NeurIPS 2019 · similarity 0.87)
- Nonparanormal Information Estimation (ICML 2017 · similarity 0.86)
- The Polynomial Stein Discrepancy for Assessing Moment Convergence (ICML 2025 · similarity 0.85)
- Stochastic Localization via Iterative Posterior Sampling (ICML 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.
