arXiv Machine Learning
Oct 8
Generalised Score Matching on Convex Domains
arXiv:2609. 11521v2 Announce Type: replace Abstract: Score matching avoids computing the normalising constant that maximum-likelihood estimation requires.
Generalized Score Matching for Parameter Estimation on Convex Domains
The paper introduces a generalized score matching objective for parameter estimation on convex subsets of ρ^d, derived from Minimum Probability Flow learning. It shows that this objective is a proper local scoring rule of second order, ensuring recovery of the true density when minimized, and proves convexity and consistency for exponential family models under standard conditions. Experiments demonstrate the method’s effectiveness on constrained domains where the partition function is intractable, including a generative modeling use‑case.
Related stories
arXiv Machine Learning
Sep 10
DDPM Score Matching and Distribution Learning
arXiv:2504.05161v2 Announce Type: replace-cross Abstract: Score estimation is the backbone of score-based generative models (SGMs), especially denoising diffusion probabilistic models (DDPMs). A key...
arXiv Machine Learning
Jun 15
Direct Fisher Score Estimation for Likelihood Maximization
arXiv:2506. 06542v2 Announce Type: replace-cross Abstract: We study the problem of likelihood maximization when the likelihood function is intractable but model simulations are readily available.
arXiv Machine Learning
Aug 5
Heteroscedasticity of Denoising Score Matching with Generalised Smooth Noise
arXiv:2508. 01597v2 Announce Type: replace Abstract: Score Matching (SM) is a powerful framework for estimating the log-density derivatives of a distribution without calculating its normalizing constants.
arXiv Machine Learning
Jun 19
Global Convergence of Gradient Descent for Score Matching in Gaussian Mixtures via Reverse Fisher Divergence
arXiv:2606. 19876v1 Announce Type: new Abstract: The score matching problem is a central training objective in modern generative modeling, diffusion models, fitting unnormalized statistical models, and inverse problems.
arXiv Machine Learning
Aug 10
Robust inference using density-powered Stein operators
arXiv:2511. 03963v3 Announce Type: replace-cross Abstract: We introduce a density-power weighted variant of the Stein operator, called the $\gamma$-Stein operator, for robust inference with unnormalized probability models.
arXiv AI
Sep 10
Newton Matching for Generative Modeling: A Unified Framework for Fine-Tuning and Sampling
arXiv:2609.05727v1 Announce Type: cross Abstract: We develop Newton Matching, a unified framework for fine-tuning and sampling in generative modeling. The target is $\pi\propto\mu e^{\tau r}$, where...
arXiv Statistics ML
Sep 3
Robust Bayesian Inference for Unnormalized Models with Mixed-Domain Data
The paper introduces SME-BETEL, a semiparametric Bayesian method that merges score matching estimating equations with Bayesian exponentially tilted empirical likelihood to perform inference on models with intractable normalizing constants. SME-BETEL avoids evaluating these constants and eliminates the need for learning-rate calibration, while providing consistency, asymptotic normality, and a Bernstein‑von Mises theorem that guarantees asymptotically calibrated credible sets even under model misspecification. The authors extend the framework to mixed‑domain data, enabling robust inference for doubly‑intractable models such as spatial preferential sampling, and demonstrate its effectiveness through simulations and an ozone‑monitoring application.
arXiv Machine Learning
Jul 7
Tightening the Score Matching Gap for Diffusion Models
arXiv:2607. 04442v1 Announce Type: cross Abstract: Diffusion models (DMs) are a state-of-the-art generative method to approximately sample from an unknown distribution.
arXiv Machine Learning
Jun 26
Learning from a Biased Sample
arXiv:2209. 01754v5 Announce Type: replace-cross Abstract: The empirical risk minimization approach to data-driven decision making requires access to training data drawn under the same conditions as those that will be faced when the decision rule is deployed.
arXiv Machine Learning
Oct 8
Efficient Best-of-N policy evaluation for inference-time alignment
The paper introduces a sample‑only framework for evaluating and selecting Best‑of‑N (BoN) inference‑time alignment policies without needing response likelihoods. By exploiting the order‑statistic structure of BoN, the authors express required density ratios through score‑rank probabilities that can be estimated from samples alone, and develop a doubly robust estimator (BoN‑DR) that reuses a shared auxiliary sample pool across candidate budgets. The method provides valid asymptotic inference even with reward estimator misspecification, proves efficiency, and offers two selection rules—maximizing estimated policy value or a lower confidence bound—to avoid over‑optimization and guarantee no harm, as demonstrated on synthetic data and GSM8K.
arXiv Machine Learning
5d ago
Cumulative Marginal Matching for One-Step Diffusion
The paper introduces Cumulative Marginal Matching (CMM), a new objective for training few‑step diffusion and flow‑matching models. CMM treats flow‑matching as a sequence of intermediate marginal distributions and directly optimizes the cumulative Wasserstein‑2 discrepancy between the base and few‑step models across the entire trajectory. It derives a tractable surrogate for this objective, shows that Shortcut‑CSL is a first‑order approximation of CMM, and reports improved generation quality on standard benchmarks under the same training budget.