Research map / Probabilistic and scientific modeling / Variational and generative sampling
A Variational Approach to Bayesian Phylogenetic Inference
Cheng Zhang, Frederick A. Matsen IV
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add PSP branch length parameterization (ablation) | VBPI VIMCO(10) · DS1 | variational lower bound (nats): -7112.39(1.44) -> -7111.17(0.99) | Improved |
| add PSP branch length parameterization (ablation) | VBPI VIMCO(10) · DS8 | variational lower bound (nats): -8659.45(0.65) -> -8655.55(0.40) | Improved |
| add PSP branch length parameterization (ablation) | VBPI VIMCO(10) · DS3 | variational lower bound (nats): -33736.81(0.34) -> -33736.70(0.36) | No clear effect |
| add PSP branch length parameterization (ablation) | VBPI VIMCO(10) · DS1 | KL divergence of tree topology posterior | Improved |
| increase samples K in training objective from 10 to 20 (ablation) | VBPI VIMCO(10) · DS1 (split-based parameterization) | KL divergence of tree topology posterior | Improved |
| increase samples K from 10 to 20 (ablation) | VBPI VIMCO · DS1 | variational lower bound (nats): -7112.39(1.44) -> -7112.60(2.10) | No clear effect |
| increase samples K from 10 to 20 (ablation) | VBPI VIMCO · DS1 | log marginal likelihood (nats): -7108.48(0.26) -> -7108.46(0.23) | No clear effect |
| use VBPI (VIMCO(20)+PSP) instead of stepping-stone estimator | SS baseline · DS7 | log marginal likelihood (nats): -37332.76(2.42) -> -37332.03(0.31) | Improved |
| use VBPI instead of MrBayes MCMC for tree posterior | MrBayes MCMC · DS1 | KL divergence of tree topology posterior | Improved |
| use short-run MCMC subsplit support instead of bootstrap with faster annealing (ablation) | VBPI · DS1 | convergence time and KL divergence: beta_t t/100,000 -> t/50,000 | Mixed |
| use RWS instead of VIMCO estimator (ablation) | VBPI · DS1 | KL divergence of tree topology posterior | Mixed |
| use bito library instead of Python autodiff for likelihood and gradient | VBPI · DS1 | wall time for 200,000 iterations (seconds): 14200 -> 1960 | Improved |
Closest papers by meaning
- Variational Phylogenetic Inference with Products over Bipartitions (ICML 2025 · similarity 0.93)
- Likelihood-free inference of phylogenetic tree posterior distributions (NeurIPS 2026 · similarity 0.91)
- PhyloGen: Language Model-Enhanced Phylogenetic Inference via Graph Structure Generation (NeurIPS 2024 · similarity 0.89)
- Deep Generative Models for Phylogenetic Inference with Complex Evolutionary Processes (NeurIPS 2026 · similarity 0.88)
- PhyloGFN: Phylogenetic inference with generative flow networks (ICLR 2024 · similarity 0.87)
- PhyloVAE: Unsupervised Learning of Phylogenetic Trees via Variational Autoencoders (ICLR 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
