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..
Restricted Spectral Gap Decomposition for Simulated Tempering Targeting Mixture Distributions
Authors: Jhanvi Garg, Krishnakumar Balasubramanian, Quan Zhou
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | By working with the restricted spectral gap, the applicability of our results is extended to broader settings such as when the usual spectral gap is difficult to bound or becomes degenerate. We demonstrate the application of our theoretical results by analyzing simulated tempering combined with random walk Metropolis Hastings for sampling from mixtures of Gaussian distributions. Our complexity bound scales polynomially with the separation between modes, logarithmically with 1/ε, where ε denotes the target accuracy in total variation distance, and exponentially with the dimension d. ... In Section 4, we empirically validate our convergence guarantee for the STMH algorithm by sampling from a two-dimensional Gaussian mixture. |
| Researcher Affiliation | Academia | Jhanvi Garg Department of Statistics Texas A&M University College Station, TX 77843 EMAIL Krishnakumar Balasubramanian Department of Statistics University of California Davis, CA 95616 EMAIL Quan Zhou Department of Statistics Texas A&M University College Station, TX 77843 EMAIL |
| Pseudocode | Yes | Algorithm 1 Simulated Tempering Metropolis Hastings Algorithm Algorithm 2 Partition Function Estimation |
| Open Source Code | Yes | Justification: The code is included as a ZIP file (code.zip) to reproduce the experiments. |
| Open Datasets | No | To numerically investigate the complexity of the STMH algorithm, we perform a simulation study with target distribution being a symmetric two-dimensional Gaussian mixture distribution, whose density is given by... This paper does not introduce or release any new assets. |
| Dataset Splits | No | To numerically investigate the complexity of the STMH algorithm, we perform a simulation study with target distribution being a symmetric two-dimensional Gaussian mixture distribution, whose density is given by... The paper does not involve standard training/test splits. However, all algorithmic parameters are clearly specified. |
| Hardware Specification | Yes | All simulations were performed on a standard consumer-grade CPU with parallelization and completed within approximately six hours. |
| Software Dependencies | No | The paper does not explicitly mention specific software dependencies with version numbers. |
| Experiment Setup | Yes | For each value of D, we run the STMH algorithm with the parameters specified in Appendix C.4.1 and initialized at (10, 10). To assess convergence, we monitor how quickly the empirical mean of the samples, denoted by ˆµ, approaches the true mean of the target distribution, (0, 0). ... To analyze how the mixing time depends on the threshold ε, we fix D = 30 and plot how ˆµ varies with the number of steps N in the right panel of Figure 1. |