v0.1.0: original BORF (unweighted)

Freezes the original Bag-Of-Receptive-Fields (BORF) implementation for regularly sampled time series. Later releases replace this API.

Install

pip install "git+https://github.com/fspinna/borf@v0.1.0"

Tested with

Python 3.12, scikit-learn 1.6.1, NumPy 1.26.4.

scikit-learn 1.6 warns that transforming an unfitted Pipeline "will raise an error in 1.8", so newer scikit-learn versions are expected to fail. Pin it:

pip install "scikit-learn<1.8"

NumPy 2 has not been tested.

Usage

See the README at this tag and notebooks/xai_example.ipynb for explanations.