prob-conf-mat 0.4.0

Confusion matrices with uncertainty quantification, experiment aggregation and significance testing.

Probabilistic Confusion Matrices

prob_conf_mat is a Python package for performing statistical inference with confusion matrices. It quantifies the amount of uncertainty present, aggregates semantically related experiments into experiment groups, and compares experiments against each other for significance.

Installation

Installation can be done using from pypi can be done using pip:

pip install prob_conf_mat

Or, if you're using uv, simply run:

uv add prob_conf_mat

The project currently depends on the following packages:

Dependency tree

prob-conf-mat
├── jaxtyping
├── matplotlib
├── numpy
├── scipy
└── tabulate

Additionally, pandas is an optional dependency for some reporting functions.

Development Environment

This project was developed using uv. To install the development environment, simply clone this github repo:

git clone https://github.com/ioverho/prob_conf_mat.git

And then run the uv sync --dev command:

uv sync --dev

The development dependencies should automatically install into the .venv folder.

Quick Start

from prob_conf_mat import Study

study = Study(seed=0, num_samples=10000, ci_probability=0.95)

study.add_experiment(
  experiment_name="baseline",
  confusion_matrix=[[50, 2, 1], [3, 45, 4], [0, 5, 40]]
)

study.add_metric("f1@macro")

print(study.report_metric_summaries(metric="f1@macro", table_fmt="github"))

Group	Experiment	Observed	Median	Mode	95.0% HDI	MU	Skew	Kurt
baseline	baseline	0.8992	0.9001	0.9043	[0.8507, 0.9448]	0.0941	-0.4361	0.3485

Rather than a single point estimate, you get a distribution of plausible macro F1 score, which we summarize using a variety of statistics.

See the Getting Started tutorials and the worked-example notebooks in docs/getting_started/examples for more.

Documentation

For more information about the package, motivation, how-to guides and implementation, please see the documentation website. We try to use Daniele Procida's structure for Python documentation.

The documentation is broadly divided into 4 sections:

1. Getting Started: a collection of small tutorials to help new users get started
2. How To: more expansive guides on how to achieve specific things
3. Reference: in-depth information about how to interface with the library
4. Explanation: explanations about why things are the way they are

	Learning	Coding
Practical	Getting Started	How-To Guides
Theoretical	Explanation	Reference

Development

This project was developed using the following (amazing) tools:

1. Package management: uv
2. Linting: ruff
3. Static Type-Checking: basedpyright
4. Documentation: zensical
5. Pre-commit: prek

Most of the common development commands are included in ./Makefile. If make is installed, you can immediately run the following commands:

Usage:
  make <target>

Utility
  help             Display this help
  hello-world      Tests uv and make
  clean            Clean up caches and build artifacts

Environment
  install          Install default dependencies
  install-dev      Install dev dependencies
  upgrade          Upgrade installed dependencies
  export           Export uv to requirements.txt file

Testing, Linting, Typing & Formatting
  test             Runs all tests
  coverage         Checks test coverage
  lint             Run linting
  type             Run static typechecking
  commit           Run pre-commit checks
  commit-log       Run pre-commit checks in verbose mode and log output to external file

Documentation
  docs-build       Update the docs
  docs-serve       Serve documentation site

Credits

The following are some packages and libraries which served as inspiration for aspects of this project: arviz, bayestestR, BERTopic, jaxtyping, mici, , python-ci, statsmodels.

A lot of the approaches and methods used in this project come from published works. Some especially important works include:

1. Goutte, C., & Gaussier, E. (2005). A probabilistic interpretation of precision, recall and F-score, with implication for evaluation. In European conference on information retrieval (pp. 345-359). Berlin, Heidelberg: Springer Berlin Heidelberg.
2. Tötsch, N., & Hoffmann, D. (2021). Classifier uncertainty: evidence, potential impact, and probabilistic treatment. PeerJ Computer Science, 7, e398.
3. Kruschke, J. K. (2013). Bayesian estimation supersedes the t test. Journal of Experimental Psychology: General, 142(2), 573.
4. Makowski, D., Ben-Shachar, M. S., Chen, S. A., & Lüdecke, D. (2019). Indices of effect existence and significance in the Bayesian framework. Frontiers in psychology, 10, 2767.
5. Hill, T. (2011). Conflations of probability distributions. Transactions of the American Mathematical Society, 363(6), 3351-3372.
6. Chandler, J., Cumpston, M., Li, T., Page, M. J., & Welch, V. J. H. W. (2019). Cochrane handbook for systematic reviews of interventions. Hoboken: Wiley, 4.

Citation

@software{ioverho_prob_conf_mat,
    author = {Verhoeven, Ivo},
    license = {MIT},
    title = {{prob\_conf\_mat}},
    url = {https://github.com/ioverho/prob_conf_mat}
}