C++使用吉布斯采样器(即狄利克雷过程高斯混合模型)实现 GMM
我正在寻找一个多变量 GMM 的 C++ 实现,它使用基于吉布斯采样的方法来拟合/分类(而不是通常的基于 EM),以便能够充分利用先验信息并添加约束。通常称为狄利克雷过程高斯混合模型或 DPGMM。
我已经在 Matlab 中实现了此功能,但没有花时间转换此代码(是的,我的代码使用内置的 matlab 编码器进行转换,但它目前依赖于各种附加的 Matlab 库)。效率也很重要,我会每秒多次将 GMM 拟合到大型数据集。
因此,我有兴趣知道是否已经存在众所周知的高效代码。最初的搜索并没有得到太多结果。
I am looking a C++ implementation of a multi-variate GMM that uses a Gibbs Sampling based approach to fitting / classification (rather than the usual EM based), in order to be able to make full use of a priori information and add in constraints. Often known as a Dirichlet Process Gaussian Mixture Model or DPGMM.
I already have this implemented in Matlab, but rather than spending time converting this code (yes I code use the built in matlab coder to convert, but it currently relies on various additional Matlab libraries). Also efficiency is important, I will be fitting a GMM to large data sets many times a second.
Thus, I am interested to know if there was already well known efficient code out there. An initial search hasn't returned very much.
如果你对这篇内容有疑问,欢迎到本站社区发帖提问 参与讨论,获取更多帮助,或者扫码二维码加入 Web 技术交流群。

绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱,如果其他用户或者作者回复了您的评论,将不能在第一时间通知您!
发布评论
评论(1)
虽然不特定于 GMM,但您可以使用 CppBugs 项目来指定您自己的模型并让库运行模拟。
While not specific to GMM's you could use the CppBugs project to specify your own model and let the library run the simulation.