Probabilistic Path Hamiltonian Monte Carlo
Authors: Vu Dinh, Arman Bilge, Cheng Zhang, Frederick A. Matsen IV
ICML 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this section, we demonstrate the validity and efficiency of our PPHMC method by an application to Bayesian phylogenetic inference. We compared our PPHMC implementations to industry-standard Mr Bayes 3.2.5, which uses MCMC to sample phylogenetic trees (Ronquist et al., 2012). We first tested our PPHMC method on a simulated data set. |
| Researcher Affiliation | Academia | 1Program in Computational Biology, Fred Hutchison Cancer Research Center, Seattle, WA, USA 2Department of Statistics, University of Washington, Seattle, WA, USA. |
| Pseudocode | Yes | Algorithm 1 Leap-prog algorithm with step size ϵ. Algorithm 2 Refractive Leap-prog with surrogate |
| Open Source Code | Yes | We validate the algorithm through two independent implementations in open-source software: 1. a Scala version available at https://github.com/armanbilge/phylo HMC that uses the Phylogenetic Likelihood Library1 (Flouri et al., 2015), and 2. a Python version available at https://github. com/zcrabbit/Phylo Infer that uses the ETE toolkit (Huerta-Cepas et al., 2016) and Biopython (Cock et al., 2009). |
| Open Datasets | Yes | As a proof of concept, we first tested our PPHMC method on a simulated data set. We used a random unrooted tree with N = 50 leaves sampled from the aforementioned prior. 1000 nucleotide observations for each leaf were then generated by simulating the continuous-time Markov model along the tree. We also analyzed an empirical data set labeled DS4 by Whidden and Matsen (2015) that has become a standard benchmark for MCMC algorithms for Bayesian phylogenetics since Lakner et al. (2008). |
| Dataset Splits | No | The paper mentions burn-in periods for MCMC runs ('burn-in period of the first 25% iterations', 'burn-in of 25%'), but does not specify training, validation, or test dataset splits in terms of percentages or sample counts for model development or evaluation. |
| Hardware Specification | No | The paper does not specify any particular hardware (e.g., CPU, GPU models, memory) used for running the experiments. |
| Software Dependencies | Yes | We compared our PPHMC implementations to industry-standard Mr Bayes 3.2.5, which uses MCMC to sample phylogenetic trees (Ronquist et al., 2012). a Scala version available at https://github.com/armanbilge/phylo HMC that uses the Phylogenetic Likelihood Library1 (Flouri et al., 2015), and 2. a Python version available at https://github. com/zcrabbit/Phylo Infer that uses the ETE toolkit (Huerta-Cepas et al., 2016) and Biopython (Cock et al., 2009). |
| Experiment Setup | Yes | For PPHMC, we set the step size ϵ = 0.0015 and smoothing threshold δ = 0.003 to give an overall acceptance rate of about α = 0.68 and set the number of leap-prog steps T = 200. |