Nonparametric Involutive Markov Chain Monte Carlo
Authors: Carol Mak, Fabian Zaiser, Luke Ong
ICML 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our empirical study shows that the existing strengths of several i MCMC algorithms carry over to their nonparametric extensions. Applying our method to the recently proposed Nonparametric HMC, an instance of (Multiple Step) NP-i MCMC, we have constructed several nonparametric extensions (all of which new) that exhibit significant performance improvements. (Abstract) and 5. Experiments (Section Title) |
| Researcher Affiliation | Academia | Carol Mak 1 Fabian Zaiser 1 Luke Ong 1 1Department of Computer Science, University of Oxford, United Kingdom. Correspondence to: Carol Mak <pui.mak@cs.ox.ac.uk>. |
| Pseudocode | Yes | Listing 1. Infinite Gaussian mixture model (Page 2) and Listing 2. Pseudocode of the Hybrid NP-i MCMC algorithm (Page 11), among others. |
| Open Source Code | Yes | The code to reproduce the Turing experiments is available in https://github.com/cmaarkol/nonparametricmh. |
| Open Datasets | No | The paper mentions '30 data points generated from a ground truth with three components' for GMM and a 'training data set generated from a mixture of 9 components' for the Gaussian mixture model. However, it does not provide concrete access information (link, DOI, repository, or formal citation with authors/year) for these specific generated datasets. |
| Dataset Splits | No | The paper mentions 'training data set' and 'test data set' but does not provide specific details on validation splits, percentages, or absolute sample counts for train/validation/test sets. |
| Hardware Specification | No | The paper does not explicitly describe any specific hardware used for running its experiments (e.g., GPU/CPU models, memory, or cloud instance types). |
| Software Dependencies | No | The paper mentions implementing experiments 'in the Turing language (Ge et al., 2018)' but does not specify the version number for Turing or any other ancillary software components and libraries with their versions. |
| Experiment Setup | Yes | The results of ten runs with 5000 iterations each (Fig. 3) suggest that the NP-i MCMC samplers work pretty well. (Page 7) and we ran NP-DHMC for a step count L {2, 5} with and without persistence. (Page 7) and Each run: 10^3 samples, L leapfrog steps, step size ϵ = 0.1, persistence parameter α {0.5}. (Table 1, Page 7) and Each run: 10^3 samples with L = 5 leapfrog steps of size ϵ = 0.1, persistence parameter α {0.5, 0.1}, and look-ahead K {1, 2}. (Figure 4 caption, Page 8) and Each run: 10^3 samples with L = 25 leapfrog steps of size ϵ = 0.05, persistence parameter α = 0.5, and look-ahead K {1, 2}. (Figure 5 caption, Page 8). |