Installation

Fastchange can be installed in three different ways: conda, pip, or directly through Git. Conda is the preferred method of installation as it provides faster scipy/numpy distributions that pip, and offers the icc_rt library that can be used for an additional performance increase (see Installing with extras).

1. Via conda (preferred) conda install -c gursky1 fastchange

2. Via pip pip install fastchange

3. Via git pip install git+https://github.com/gursky1/fastchange

Installing with extras

svml

As per the numba documentation, you can get a performance increase using Intel's SVML library icc_rt. Note that this is only available via numba's conda channel:

conda install -c numba icc_rt

To install fastchange with icc_rt out of the box, install with the "svml" extra:

pip/conda install fastchange[svml]

r

Fastchange also offers an interface to several R changepoint libraries using the rpy2 package. Note that this requires a local installation of R to function. Fastchange provides integration with the changepoint, changepoint.np, and bocp R libraries. Note these also need to be manually installed to function. The R interface can be installed using the "r" extra keyword:

With conda:

conda install -c gursky1 fastchange and pip install rpy2

Or with pip: pip install fastchange[r]