hep_ml

What's Changed

• Upgrade sklearn criterion mse to squared_error by @jonas-eschle in https://github.com/arogozhnikov/hep_ml/pull/71
• Resolve conflicts between numpy 1.22 and theano, stop testing python 2.7 by @arogozhnikov in https://github.com/arogozhnikov/hep_ml/pull/72
• replace np.float with float, fix deprecation warning by @ahill187 in https://github.com/arogozhnikov/hep_ml/pull/75
• replace deprecated numpy aliases with python builtins by @richard-lane in https://github.com/arogozhnikov/hep_ml/pull/78
• remove theano from testing, as theano can't work with current numpy (… by @arogozhnikov in https://github.com/arogozhnikov/hep_ml/pull/79
• Fix creation of releases from pypi by @arogozhnikov in https://github.com/arogozhnikov/hep_ml/pull/80
• bump version by @arogozhnikov in https://github.com/arogozhnikov/hep_ml/pull/81

New Contributors

• @ahill187 made their first contribution in https://github.com/arogozhnikov/hep_ml/pull/75
• @richard-lane made their first contribution in https://github.com/arogozhnikov/hep_ml/pull/78

Full Changelog: https://github.com/arogozhnikov/hep_ml/compare/v0.7.0...v0.7.2

• fixed weight normalization (@jonas-eschle)
• moved travis ci -> github actions
• auto-deployment of new releases to pypi
• documentation was moved to /docs and served from there

• updated jupyter examples
• some minor fixes to adapt to updates in dependencies (numpy, sklearn, theano)
• updated CI scripts
• minor improcements in the documentation

Fixed problems with uboost and theano

• fixed some examples
• updated docs
• added some parameters to GBReweighter

• hep_ml.speedup (speeding up predictions) is published
• hep_ml.splot (minimalistic splot) is published
• some improvements to reweighter and losses

• awesome Gradient Boosted reweighter is published
  https://github.com/arogozhnikov/hep_ml/blob/master/notebooks/DemoReweighting.ipynb
• demonstration of fittting expression with nnet is published
  https://github.com/arogozhnikov/hep_ml/blob/master/notebooks/DemoNeuralNetworks.ipynb
• newer & fulller documentation
  https://arogozhnikov.github.io/hep_ml/
• some enhancements in UGradientBoosting and losses

Tags are required to get info on old releases

First release of minimized version.

• moved to different repository
• first public version with PyPI support
• completely rewritten gradient boosting and loss functions
• added support of neural networks
• standard interface for losses (including RankBoostLoss)
• REP interface for metrics
• first version of documentation