Xcessiv Versions Save

A web-based application for quick, scalable, and automated hyperparameter tuning and stacked ensembling in Python.

v0.5.1

6 years ago

Bugfix:

  • #51 fixes #50, a bug resulting from changes in scikit-learn v0.19

New features:

  • Docker-compose file for easier docker deployment by @marcelmaatkamp

v0.5.0

6 years ago

Features

  • #43 Automated ensembling is finally here! Greedy forward model selection introduced

Warning: Project files from version before v0.5.0 will not work with v0.5.0

v0.4.0

6 years ago

Features

  • #35 by @menglewis
  • #37 Added TPOT integration

Warning: v0.4 Project files are not compatible with earlier versions

v0.3.8

6 years ago

Features

  • #29 #30 Removed append original checkbox and added Identity Transformer preset base learner instead

v0.3.7

6 years ago

Features

  • #27 More estimators (regressors) by @enisnazif
  • #28 "Export ensemble as Python package" changed to "Export ensemble as Python file". Also, an additional shortcut for directly exporting a stacked ensemble as a base learner setup. Awesome!

v0.3.6

6 years ago

Features

  • #23 More preset cross-validators by @ryanliwag
  • #24 Added preset metrics median absolute error, R2 score, and explained variance score
  • #26 Added functionality that stores previous parameter searches. More of a user experience fix.

v0.3.5

6 years ago

New Feature:

  • #22 Added ability to export a stacked ensemble as a Python package so you can use it on different data.
  • #19 by @jef5ez adds Mean Absolute Error as preset metric

Docs:

  • #22 Added docs for using exported stacked ensemble Python package
  • #20 Added docs for using TPOT with Xcessiv

v0.3.4

6 years ago

Hotfix for 0.3.0. Ended up at 0.3.4 because Pypi problems..

v0.3.0

6 years ago

Major feature addition

  • #18 - Added an experimental bayesian optimization search beside grid search and random search to allow a bit of automation for hyperparameter tuning.

v0.2.5

6 years ago

Hotfix for Python 3 users.