使用 Homebrew 安装 libbi

项目链接

• 主页
• 仓库
• 上游文档
• 软件包管理器页面

命令和别名

• bi
• libbi

项目历史

LibBi was publicly announced on 2013-06-07, alongside Murray's 2013 paper on Bayesian state-space modelling on high-performance hardware. The paper presents LibBi as software that parses a model language, optimizes it, generates C++ code, compiles it, and runs inference methods on CPU, GPU, and distributed-memory platforms.

The project site frames LibBi around sequential Monte Carlo methods, including particle filtering, PMCMC, SMC^2, the extended Kalman filter, and parameter optimization routines. Its design reflects a scientific-computing period when CUDA, OpenMP, MPI, NetCDF, and HDF5 were expected to be stitched together by domain-specific tooling.

采用历史

LibBi found a niche in statistical and scientific modelling rather than general-purpose application development. The project site records an RBi package announcement on 2016-10-19 for using LibBi from R, and a 2016-11-14 post announcing Homebrew and Linuxbrew installation for `libbi`.

使用方式

Users write state-space models in LibBi's modelling language, then run the `libbi` or `bi` tools to perform inference and process NetCDF/HDF5-backed input and output. The package is aimed at researchers who want particle methods and hardware parallelism without hand-writing the full generated C++ implementation.

为什么软件包爱好者会关心

LibBi is a good example of a package-manager formula preserving a research software stack: C++, Perl, CUDA-era assumptions, NetCDF/HDF5-style data, and a custom DSL all bundled as a command-line tool. It is niche, but interesting because the package is the executable surface of an academic modelling system.

时间线

• 2013: Initial release of LibBi is announced on 2013-06-07.
• 2013: Introductory LibBi paper is listed by the project for citation.
• 2016: RBi interface for R is announced on 2016-10-19.
• 2016: Homebrew and Linuxbrew installation is announced on 2016-11-14.
• 2019: Project news lists LibBi 1.4.5 on 2019-07-08.

Related projects

• RBi, R, MATLAB, GNU Octave, Julia, CUDA, OpenMP, MPI, NetCDF, HDF5, and sequential Monte Carlo research code are the most relevant neighbors.

来源

• https://github.com/lawmurray/LibBi
• https://libbi.org/
• https://libbi.org/2013/06/07/LibBi-Released
• https://libbi.org/2016/10/19/LibBi-package-for-R
• https://libbi.org/2016/11/14/Easily-install-LibBi-from-Homebrew
• https://libbi.org/documentation
• https://libbi.org/getting-started
• https://libbi.org/papers