Parameter elimination in particle Gibbs sampling
Authors: Anna Wigren, Riccardo Sven Risuleo, Lawrence Murray, Fredrik Lindsten
NeurIPS 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We have implemented PG, m PG, PGAS and m PGAS in Birch [24] and provide examples to illustrate their efficiency in Section 4.2 and 4.3. |
| Researcher Affiliation | Collaboration | Anna Wigren Department of Information Technology Uppsala University, Sweden anna.wigren@it.uu.se Riccardo Sven Risuleo Department of Information Technology Uppsala University, Sweden riccardo.risuleo@it.uu.se Lawrence Murray Uber AI San Francisco, CA, USA lawrence.murray@uber.com Fredrik Lindsten Division of Statistics and Machine Learning Linköping University, Sweden fredrik.lindsten@liu.se |
| Pseudocode | Yes | Algorithm 1 SMC; Algorithm 2 Marginalized PGAS for the restricted exponential family; Algorithm 3 Blocking for m PG/m PGAS; Algorithm 4 Marginalized particle marginal Metropolis Hastings |
| Open Source Code | Yes | Code for all numerical simulations is available at https://github.com/uu-sml/neurips2019-parameter-elimination. |
| Open Datasets | No | The paper mentions datasets like "a data set from an outbreak on the island of Yap in Micronesia in 2011" and "a dataset of observations of the number of song sparrows on Mandarte Island, British Columbia, Canada [33]". However, for the first, no access information is provided. For the second, it cites a paper [33] that describes the data, but this citation does not provide direct access to the dataset itself, nor does the paper provide a link to the dataset. |
| Dataset Splits | No | The paper discusses simulation runs, burn-in periods, and number of particles, but it does not specify explicit training/test/validation dataset splits with percentages, sample counts, or citations to predefined splits. |
| Hardware Specification | No | The paper mentions that some results were obtained from an implementation in Matlab and notes computational times using Matlab's tic-toc timer, but it does not provide any specific hardware details such as CPU/GPU models, processor types, or memory specifications. |
| Software Dependencies | No | The paper mentions using "Matlab" and that they "implemented PG, PGAS, m PG and m PGAS in Birch [24]". However, it does not provide specific version numbers for either Matlab or Birch, which are necessary for reproducible software dependencies. |
| Experiment Setup | Yes | We initialized with σ2 v = σ2 w = 100 and used a bootstrap proposal for PGAS and a marginalized bootstrap proposal for m PGAS. PGAS and m PGAS were run for M = 10000 iterations, discarding the first 1500 samples as burn-in. We used B = 5 and L = 20, all other settings were the same as before. In Figure 5 (left), we report the histogram of the distribution of the density regulation parameter c estimated using 10000 samples drawn using Algorithm 4 after a burn-in of 5000 samples, using N = 512 particles. |