Releases: tokebe/niclassify
Releases · tokebe/niclassify
1.5.4 beta
- Fixed a bug where GBIF results without a
type
property would cause status lookup failure
Full Changelog: v1.5.3-beta...v1.5.4-beta
1.5.3 beta
- Changed scikit-learn version to 0.24.0
- Fixed a typo in the manual
- Updated Acknowledgments
1.5.1 Beta
- Fixed a bug which would cause the program to crash upon finding stop codons while checking reading frame
- The classifier now uses
balanced_subsample
instead ofbalanced
for class weighting, which should slightly improve training quality when training on slightly unbalanced data.
1.5.0 Beta
- When training the classifier, instead of removing any rows with any missing data, columns and rows with all missing data are removed.
- When rows/columns are removed, this is logged and the user is notified.
- lxml upgraded to latest security patch
1.4.0 Beta
1.3.0 Beta
- User may now save and load species delimitations, meaning that the user can supply their own delimitation and skip that computational step.
- File reading will read csvs/tsvs/etc successfully regardless of improper file typing.
- Removed deprecated files
- Default number of classifiers changed to 100 to cut default train time and avoid extreme overfits
- Feature generation now logs properly
- Fixed various minor bugs and message typos
1.2.0 Beta
- Split the work of fasta alignment by a given taxonomic level. This should be faster and improve the quality of alignments (fewer fixes for the user to make).
- The program now detects any uncaught exceptions and saves a traceback file, alerting the user.
1.1.2 Beta
- Fix #17
1.1.1 Beta
- Require latest Pillow to fix Pillow-related vulnerabilities
1.1.0 Beta
- Support for Linux
- Automated installation (easy launch) support for Ubuntu Linux
- Experimental Support for Mac (automated installation not yet final)
- Fix downloading of large files from BOLD
- Various package updates to address url-access vulnerabilities