VaiPhy: a Variational Inference Based Algorithm for Phylogeny

Authors: Hazal Koptagel, Oskar Kviman, Harald Melin, Negar Safinianaini, Jens Lagergren

NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental 5 Experiments, In Table 1, we provide the mean LL scores and standard deviations. On all datasets except DS2, our ϕ-CSMC is the superior CSMC method (highlighted in red). The wall-clock time comparison of the methods on DS1 is presented in Fig. 3.
Researcher Affiliation Academia 1School of EECS, KTH Royal Institute of Technology, Stockholm, Sweden 2Science for Life Laboratory, Solna, Sweden
Pseudocode Yes Alg. 1 is a high-level algorithmic description of Vai Phy; In Alg. 2, we summarize the JC sampler with an algorithmic description.
Open Source Code Yes We provide our code on Git Hub: https://github.com/Lagergren-Lab/Vai Phy.
Open Datasets Yes Here we benchmark our methods, Vai Phy, and ϕ-CSMC, in terms of LL estimates on seven real-world datasets, which we refer to as DS1-DS7 ([16, 13, 35, 17, 25, 40, 29]; in Appendix E, we provide additional information about the datasets).
Dataset Splits No No specific training/validation/test dataset splits (percentages or sample counts) are mentioned in the paper's main text.
Hardware Specification Yes The experiments are performed on a high-performance computing cluster node with two Intel Xeon Gold 6130 CPUs with 16 CPU cores each. Each node in the cluster has 96 Gi B RAM.
Software Dependencies No No specific version numbers for software dependencies (e.g., Python libraries, frameworks, or specialized solvers) are provided in the paper.
Experiment Setup Yes For all methods, we use the exact parameter configurations reported in the corresponding papers or specified in their available code. Following [26], we give all CSMC methods K = 2048 particles. We run Vai Phy and ϕ-CSMC for 200 iterations and evaluate them on Eq. (16) and Eq. (17), respectively.