De-randomizing MCMC dynamics with the diffusion Stein operator
Authors: Zheyang Shen, Markus Heinonen, Samuel Kaski
NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We demonstrate empirically that GSVGD can de-randomize complicated MCMC dynamics, which combine the advantages of auxiliary momentum variables and Riemannian structure, while maintaining the high sample quality from an interacting particle system. |
| Researcher Affiliation | Academia | 1Helsinki Institute for Information Technology, HIIT, Department of Computer Science, Aalto University, Finland 2Department of Computer Science, University of Manchester |
| Pseudocode | No | The paper provides mathematical equations for particle updates (e.g., Equation 18), but does not include a clearly labeled pseudocode or algorithm block. |
| Open Source Code | No | The paper does not provide an explicit statement or link to open-source code for the described methodology. |
| Open Datasets | Yes | The test log-likelihood (higher is better) results for selected datasets in UCI repository. The results are reported in mean (standard deviation) form averaged over 20 runs in the first 6 columns and 6 runs in the last 2, with the best performing model marked in boldface. |
| Dataset Splits | No | The paper mentions running experiments on UCI datasets and reporting test log-likelihoods, but does not specify the training/validation/test splits (e.g., percentages or counts) used for these datasets. |
| Hardware Specification | No | The paper acknowledges computational resources provided by 'Aalto Science-IT Project from Computer Science IT and CSC IT Center for Science, Finland', but does not specify any particular hardware models (e.g., CPU, GPU, or memory details). |
| Software Dependencies | No | The paper does not provide specific software names with version numbers required to reproduce the experiments. |
| Experiment Setup | Yes | We apply GSVGD of advanced MCMC dynamics to the inference of Bayesian neural networks, taking a simple structure of one hidden layer, and an output with Gaussian likelihood. We opt for the fully Bayesian specification of BNN, where the precision parameter of the weight prior and the Gaussian likelihood follows Gamma(1, 0.1). |