lir 1.3.3

v1.3.3

on Python PyPI

What's Changed

Additions

• added head argument to CSV parser to read the first n rows only by @wowtor in #302
• Subset aggregation #286
• Create a YAML-schema for validation and development goals #304
• Add mcmc simulations #316
• Add EuclideanDistance #310

Paired data

• Allow Predefined Train Tests splits for paired instances #284
• Csv parser supports reading pair data by reading two source id columns #282
• Composite transformer that fits separate instances for categories #283

Improvements

• Properly close plots #291
• on error, do not print the full stack trace except at log level DEBUG #268
• refactor IsotonicCalibrator to implement the Transformer interface #289
• Improve comments & documentation #298 #290 #305 #319 #320
• Allow launching lir directly as a module: python -m lir, instead of python -m lir.main #309
• Move experiments to dedicated experiments module #308
• Move datasets from lir/data/datasets to lir/datasets #307
• Remove last linting exceptions / no longer require Ax object when plotting. #303
• added all attribute to some of the modules #318

Fixes

• Update synthesized binary data generator, to work with the latest FeatureData class by @wowtor in #288
• devpav now returns a single float, not an array with one float by @wowtor in #296
• SaveModel aggregation now uses run_name in the target path by @wowtor in #295
• Replaced autodoc by autodoc2 and parse docstr as markdown by @wowtor in #297
• cllr_min no longer returns NaN on invalid input (like cllr) by @wowtor in #299
• metrics aggregation continues even if a metric cannot be calculated by @wowtor in #300
• IsotonicRegression can handle NaN #312

Full Changelog: v1.3.2...v1.3.3