Hidden Markov models in PyMC3.
- Fully implemented PyMC3
Distribution
classes for HMM state sequences and mixtures that depend on them - A forward-filtering backward-sampling (FFBS) implementation that works with NUTS—or any other PyMC3 sampler
- A conjugate Dirichlet transition matrix sampler
- Support for time-varying transition matrices in both the
Distribution
classes and FFBS sampler
The package name is pymc3_hmm
and it can be installed with pip
directly from GitHub
$ pip install git+https://github.com/nccmedia/pymc3-hmm
First, pull in the source from GitHub:
$ git clone git@github.com:NCCMedia/pymc3-hmm.git
Afterward, you can run make conda
or make venv
to set up a virtual environment. After making changes, be sure to run make black
in order to automatically format the code and then make check
to run the linters and tests.