optuna 0.5.0

v0.5.0

on Python PyPI

This is the release note of v0.5.0. See here for the complete list of solved issues and merged PRs.

Bug Fixes

• Make samplers/tpe/parzen_estimator.py adaptive. (#261, thanks @HideakiImamura!)

Tests

• Add test cases to check NaN handling of MedianPruner. (#264)

New Features

• Add FixedTrial. (#250)
• Add load_if_exists option. (#242)

Examples

• Fix a typo in pruning example of xgboost. (#266)
• Fix an invalid parameter, ets, in xgboost example. (#265, thanks @higumachan!)
• Use list for choice argument instead of tuple. (#252)

Documents

• Fix a style error in FAQ. (#267)
• Add an FAQ about the way to specify the GPU with CUDA_VISIBLE_DEVICES. (#263)
• Add an FAQ about NaN returned by objective functions. (#258)
• Add an FAQ about the way to save machine learning models. (#255)
• Add an FAQ about random seed. (#253)
• Add an FAQ about logging levels. (#251)

Enhancements

• Fix typo. (#269, thanks @RossyWhite!)
• Overwrite sigma of prior distributions. (#268, thanks @fofof200!)
• Make TYPE_CHECKING work before Python 3.5.2. (#262, thanks @okapies!)
• Make ChainerPruningExtension report NaN to pruner. (#259)
• Fix mypy's version constraint. (#257, thanks @okapies!)