Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
Nearly Dimension-Independent Convergence of Mean-Field Black-Box Variational Inference
Authors: Kyurae Kim, Yian Ma, Trevor Campbell, Jacob Gardner
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Theoretical | We prove that, given a mean-field location-scale variational family, black-box variational inference (BBVI) with the reparametrization gradient converges at a rate that is nearly independent of any explicit dimension dependence. Specifically, for a d-dimensional strongly log-concave and log-smooth target, the number of iterations for BBVI with a sub-Gaussian family to obtain a solution ϵ-close to the global optimum has an explicit dimension dependence no larger than O(log d). This is a significant improvement over the O(d) dependence of full-rank locationscale families. For heavy-tailed families, we prove a weaker O(d2/k) dependence, where k is the number of finite moments of the family. Additionally, if the Hessian of the target log-density is constant, the complexity is free of any explicit dimension dependence. |
| Researcher Affiliation | Academia | Kyurae Kim University of Pennsylvania EMAIL Yi-An Ma University of California San Diego EMAIL Trevor Campbell University of British Columbia EMAIL Jacob R. Gardner University of Pennsylvania EMAIL |
| Pseudocode | No | The paper discusses the theoretical aspects of Black-Box Variational Inference (BBVI) and stochastic proximal gradient descent (SPGD), including their setup and convergence properties, but it does not present any structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper mentions that BBVI is implemented in Stan, PyMC, Pyro, and Turing, but it does not state that the authors are releasing their own code for the methodology described in this paper, nor does it provide any links to a code repository. |
| Open Datasets | No | The paper is a theoretical work focusing on convergence guarantees for variational inference. It does not describe or use any datasets for empirical evaluation, hence no information about publicly available datasets is provided. |
| Dataset Splits | No | This paper is theoretical and does not involve experimental evaluation on datasets. Therefore, there is no mention of dataset splits (training, validation, test) or their methodologies. |
| Hardware Specification | No | The paper is purely theoretical and does not describe any experimental results. Consequently, there is no mention of specific hardware used for running experiments. |
| Software Dependencies | No | The paper is a theoretical study of convergence rates for variational inference algorithms. It does not describe an implementation of its methods or report experimental results, and thus does not list any specific software dependencies with version numbers. |
| Experiment Setup | No | The paper is theoretical and focuses on proving convergence guarantees. It does not present any experimental results, and therefore, no experimental setup details like hyperparameters or training configurations are provided. |