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 [1].
Pseudo-Marginal Hamiltonian Monte Carlo
Authors: Johan Alenlöv, Arnoud Doucet, Fredrik Lindsten
JMLR 2021 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We demonstrate through experiments that PM-HMC can outperform significantly both standard HMC and pseudo-marginal MH schemes. [...] Finally we present numerical results demonstrating the usefulness of our algorithm in Section 4. [...] We illustrate the proposed PM-HMC method on three synthetic examples and one real-world data set. |
| Researcher Affiliation | Academia | Johan Alenl ov EMAIL Division of Statistics and Machine Learning Link oping University Link oping, 581 83, Sweden Arnaud Doucet EMAIL Department of Statistics University of Oxford Oxford, OX1 3TG, United Kingdom Fredrik Lindsten EMAIL Division of Statistics and Machine Learning Link oping University Link oping, 581 83, Sweden |
| Pseudocode | Yes | Algorithm 1 Pseudo-marginal HMC (one iteration) Let (θ, u) be the current state of the Markov chain. Do: 1. Sample auxiliary variables ρ N(0d, Id) and p N(0D, ID). 2. Compute (θ , ρ , u , p ) = bΦh L(θ, ρ, u, p) using the numerical integrator (19). 3. Accept (θ , u ) with probability 1 exp(H(θ, ρ, u, p) H(θ , ρ , u , p )). |
| Open Source Code | No | The paper does not provide an explicit statement about releasing its source code or a link to a code repository. It only mentions implementing an example but not releasing the code for the main methodology. |
| Open Datasets | Yes | As a final example we consider a version of the generalized linear mixed model from Section 4.3 applied to a real-world data set. This data set is a subset of a cohort study of 275 Indonesian preschool children with 1200 observations. It has previously been studied by Zeger and Karim (1991) using Bayesian mixed models and Schmon et al. (2021) with the model used here. |
| Dataset Splits | No | The paper uses a real-world dataset but does not specify any training, validation, or test splits. The description of the experiment in Section 4.4 implies inference on the full dataset rather than a split evaluation. |
| Hardware Specification | Yes | In Figure 17 we see the average time for one iteration of the algorithm in seconds for the different methods for the different values of N used in the simulations, the simulations were done on an i7-6700k running at 4.0GHz with 16GB of memory, no GPU was used to accelerate the computations. |
| Software Dependencies | No | The paper does not provide specific software names with version numbers for the dependencies used in the experiments. |
| Experiment Setup | Yes | We apply the PM-HMC algorithm with N = 2i for i from 0 to 13. First we check the convergence of the trajectories from the numerical integrator ˆΦh L towards the trajectories for the ideal HMC algorithm. We do this by using h = 0.1 and L = 10 and look at the maximal position error over the integration period. [...] For the PM-HMC algorithm we use L = 50 leapfrog steps with a stepsize h = 0.01. |