scikit-rca 0.1.0.post1.dev0

Reliability Component Analysis

Reliability Component Analysis

Implementation of Reliability Component Analysis (RCA) from the paper

(citation pending).

scikit_rca is a scikit-learn compatible extension. The implementation is based on the scikit-learn-contrib template.

Installation

The package is available on PyPI and can be installed with:

pip install scikit-rca

Usage

Example usage is demonstrated in examples/run_rca.py. To use this example, clone the git repo and start by running

python examples/run_rca.py --help

The run_rca.py script expects to finds the data stored in the directory indicated by the --data-dir flag in the form of two files: features.npy and labels.npy, which are expected to be numpy arrays of shape [num_samples, d] and [num_samples, 2]. The labels array should be structured so that the first dimension indexes samples by group, and the second dimension provides an index of each sample within each group. An example invocation of the script is as follows:

python examples/run_rca.py \
    --data-dir /path/to/my/data \
    --lr 0.005 \
    --epochs 200 \
    --dim 5 \
    --penalty-scale 0.1 \
    --batch-size 4000 \
    --weight-decay 0.001 \
    --out-dir /path/to/store/model

Authors

Anastasia Borovykh, Max Weissenbacher, Stephanie Noble, Maxwell Shinn.