Releases: RodentDataAnalytics/mwm-ml-gen
Releases · RodentDataAnalytics/mwm-ml-gen
Version 4.0.3-beta
Bugs and issues fixed
-
Normalized feature values of NaN now equal to 0. NaN values may be generated when both min and max values of a feature are 0.
-
Fixed a hard coded variable in features computation.
-
Fixed the Labelling Quality procedure which was causing an error upon execution.
-
Fixed an issue with the agreement matrix where the numbers were generated into wrong positions.
-
Updated code for the Demo procedure.
Version 4.0.2-beta
- Fixed a bug when generating statistics
- Fixed some MAC compatibility issues
Version 4.0.1-beta
Major Changes
A number of new functionalities are available:
- Ability to label whole trajectories
- Ability to turn non-segmented trajectories into segments
- Ability to add custom class tags
- Ability to load custom computed trajectory features
- More analysis results including confidence intervals and classification statistics.
- Raw classification results are now exported for further custom analysis (or in case the Friedman test cannot be performed).
Minor Changes
- Fixed axes labels and file names.
- Better figure graphics.
- Some parts of the code have been re-written.
- Ability to read UTF16LE txt files.
Bugs and issues fixed
- New more robust classification methodology (see issue).
Version 3.1-beta
Synopsis
Major Changes
- Improved and more generalized formula for mapping segment classes to trajectories (see Gehring, Tiago V., et al. supplementary material 1.6)
- More statistics can now be computed about the classification quality and agreement of the classifiers after the majority voting merging (see wiki Class Statistics).
Minor Changes
- Browse Trajectory Data now contains a counter for the labels provided.
Known Issues
- Sometimes the user may need to press the buttons of the main GUI two times.
- The MAC release does not include the console window.
- The MAC release cannot read and parse EXCEL files. Users are advised to use TXT and CSV file format on MACs.
Version 3.1.0-test
This is a test release of version 3.1.0. Only the source code (MATLAB) is provided.
Version 3.0-beta
Major Changes
- Redesigned software and code.
- Updated classification procedure.
- New graphical user interface which includes an updated way of parsing info from exported Ethovision files.
- All the results are extracted as CSV files
- Both Windows and MAC installer.
Minor Changes
- Trajectory track: full file name is used as the track.
- All the generated CSV files have headers.
Known Issues
- Sometimes the user may need to press the buttons of the main GUI two times.
- The MAC release does not include the console window.
- The MAC release cannot read and parse EXCEL files. Users are advised to use TXT and CSV file format on MACs.
Version 2.0-beta
Major Changes
- The program can now read TXT files apart from XLSX and CSV.
- The user now has the ability to specify the generated figures' properties (font, text size, line width) and the figures' export format.
- Both Windows and MAC installer.
Minor Changes
- Some more Waitbars are added.
- p-values, confusion matrix and transition probabilities are now exported automatically to a TXT files.
Known Issues
- The MAC release does not include the console window. The user needs to see the generated text files for the numerical values (p-values, confusion matrix, transition probabilities).
- The MAC release cannot read and parse EXCEL files. Users are advised to use TXT and CSV file format on MACs.
Initial Update
Still in beta testing.
Major Changes:
- The program can now read XLSX files apart from CSV.
- Number of Clusters can now be identified with the help of a small GUI.
Minor Changes:
- The segmentation objects can now be transferred and loaded to any machine (the user may be asked to change the output folder).
- Waitbars are introduced and appear during specific processing stages (data parsing and feature computation) to indicate that the program is busy. During these stages the main GUI become invisible.
- Error messages are more detailed.
Initial Release
Still in beta testing.