https://github.com/cbg-ethz/sgs
Inference in Bayesian Networks with R
Science Score: 23.0%
This score indicates how likely this project is to be science-related based on various indicators:
-
○CITATION.cff file
-
✓codemeta.json file
Found codemeta.json file -
○.zenodo.json file
-
○DOI references
-
✓Academic publication links
Links to: arxiv.org -
○Committers with academic emails
-
○Institutional organization owner
-
○JOSS paper metadata
-
○Scientific vocabulary similarity
Low similarity (16.5%) to scientific vocabulary
Keywords
Repository
Inference in Bayesian Networks with R
Basic Info
- Host: GitHub
- Owner: cbg-ethz
- License: gpl-3.0
- Language: R
- Default Branch: master
- Homepage: https://arxiv.org/abs/2112.09217
- Size: 5.21 MB
Statistics
- Stars: 11
- Watchers: 4
- Forks: 2
- Open Issues: 0
- Releases: 0
Topics
Metadata Files
ReadMe.md
Inference in Bayesian Networks
SGS is an R package for inference in Bayesian networks. It allows for efficient exact and approximate inference that works both in low- and high-dimensional settings. Efficient marginalization is reached by splitting the calculation into sub-calculations of lower dimensionality.
This code is an implementation of the paper High-Dimensional Inference in Bayesian Networks, arXiv preprint.
Implemented exact inference methods: - SubGroupSeparation (fastest) - Junction-tree algorithm - Complete enumeration
Implemented approximate inference methods: - SubGroupSeparation (highest accuracy) - Loopy belief propagation - Markov chain Monte Carlo (MCMC) sampling
Installation
In order to install the package, it suffices to launch
R CMD INSTALL path/to/SGS
from a terminal, or make install from within the package source folder.
Being hosted on GitHub, it is possible to use the install_github
tool from an R session:
``` if (!requireNamespace("BiocManager", quietly = TRUE)) install.packages("BiocManager") BiocManager::install(c("graph", "Rgraphviz", "RBGL"))
library("devtools") install_github("cbg-ethz/SGS") ```
The packages "graph", "Rgraphviz" and "RBGL" need to be installed from BioConductor, as they are not hosted on CRAN.
SGS requires R >= 3.5, and depends on
bitops and
methods. Other packages are requested in
order to plot graphs, but are not mandatory.
Examples
``` library(SGS)
create BN and label variables
set.seed(6) myBayesNet <- randomBN(3) myBayesNet@variables <- c("rain", "sprinkler", "wet grass") plot_bn(myBayesNet)
what's the probability of having rain and wet grass at the same time?
define observed variables and calculate marginal probability
myObserved <- list(observed.vars=c("rain", "wet grass"), observed.vals=c(2,2)) exactInference(myBayesNet,myObserved)
another example:
let's learn the Bayesian network from the "Asia dataset"
asiabn <- learnbn(Asia) plotbn(asiabn)
now we can do the inference on the learned Bayesian network
myObserved <- list(observed.vars=c("X", "D"), observed.vals=c(1,1)) exactInference(asia_bn, myObserved) ```
Benchmark Results
We benchmarked the performance of our SGS method against standard inference schemes (Gibbs sampling and loopy belief propagation) over a broad range of different Bayesian networks. The results are summarized in the Figure below (lower is better), displaying the normalized root mean squared error (NRMSE). To reproduce the results, run the scripts in the benchmark folder.
Reference
If you find this code useful, please consider citing:
Fritz M. Bayer, Giusi Moffa, Niko Beerenwinkel, Jack Kuipers. High-Dimensional Inference in Bayesian Networks, arXiv preprint, 2021
@article{bayer2021marginalization,
title={High-Dimensional Inference in Bayesian Networks},
author={Bayer, Fritz M and Moffa, Giusi and Beerenwinkel, Niko and Kuipers, Jack},
journal={arXiv preprint arXiv:2112.09217},
year={2021}
}
Owner
- Name: Computational Biology Group (CBG)
- Login: cbg-ethz
- Kind: organization
- Location: Basel, Switzerland
- Website: https://www.bsse.ethz.ch/cbg
- Twitter: cbg_ethz
- Repositories: 91
- Profile: https://github.com/cbg-ethz
Beerenwinkel Lab at ETH Zurich
GitHub Events
Total
- Watch event: 1
- Fork event: 1
Last Year
- Watch event: 1
- Fork event: 1
Committers
Last synced: 10 months ago
Top Committers
| Name | Commits | |
|---|---|---|
| Fritz Bayer | f****r@y****m | 112 |
| PhysFritz | 3****z | 3 |
Committer Domains (Top 20 + Academic)
Issues and Pull Requests
Last synced: 10 months ago
All Time
- Total issues: 0
- Total pull requests: 1
- Average time to close issues: N/A
- Average time to close pull requests: less than a minute
- Total issue authors: 0
- Total pull request authors: 1
- Average comments per issue: 0
- Average comments per pull request: 0.0
- Merged pull requests: 1
- Bot issues: 0
- Bot pull requests: 0
Past Year
- Issues: 0
- Pull requests: 0
- Average time to close issues: N/A
- Average time to close pull requests: N/A
- Issue authors: 0
- Pull request authors: 0
- Average comments per issue: 0
- Average comments per pull request: 0
- Merged pull requests: 0
- Bot issues: 0
- Bot pull requests: 0
Top Authors
Issue Authors
Pull Request Authors
- fritzbayer (1)
Top Labels
Issue Labels
Pull Request Labels
Dependencies
- R >= 3.5.0 depends
- bitops * depends
- methods * depends
- Bestie * imports
- BiDAG * imports
- RColorBrewer * imports
- cowplot * imports
- ggplot2 * imports
- ggpubr * imports
- gridExtra * imports
- igraph * imports
- pcalg * imports
- stats * imports
- RBGL * suggests
- Rgraphviz * suggests
- graph * suggests
- knitr * suggests
- qgraph * suggests
- rmarkdown * suggests
- testthat * suggests
- JamesIves/github-pages-deploy-action 4.1.4 composite
- actions/checkout v2 composite
- r-lib/actions/setup-pandoc v2 composite
- r-lib/actions/setup-r v2 composite
- r-lib/actions/setup-r-dependencies v2 composite