Skip to content

Latest commit

 

History

History
195 lines (171 loc) · 10.4 KB

CHANGELOG.md

File metadata and controls

195 lines (171 loc) · 10.4 KB
Raw

Changelog

Version 0.10.2 - March 06, 2024

Changes

  • BaseProtocol.add_salvia_data now also accepts absolute saliva sample times as argument by setting sample_times_absolute=True.
  • bp.protocols.plotting.saliva_plot: removed hardcoded number of plotting colors (2) when plotting data with multiple conditions (> 2)

Version 0.10.1 - Dec 01, 2023

Bugfixes

  • biopsykit.classification.model_selection.SklearnPipelinePermuter: fixed bug that lead to the incorrect flattening of numpy arrays not due to a newer numpy version

Version 0.10.0 - Nov 28, 2023

Changes

  • biopsykit.classification.model_selection.SklearnPipelinePermuter:
    • renamed method best_pipeline to best_hyperparameter_pipeline. WARNING: This is a breaking change!
    • added new method to fit classification pipelines and save intermediate results to disk (fit_and_save_intermediate).
    • all arguments passed to fit and fit_and_save_intermediate (except for X and y) are from now on keyword-only arguments to avoid confusion. WARNING: This is a breaking change!
    • merge_permuter_instances is now a class method that accepts a list of SklearnPipelinePermuter instances or a list of file paths as argument instead of a method that merges one SklearnPipelinePermuter instance with the current instance. WARNING: This is a breaking change!
    • added new method update_permuter to update an SklearnPipelinePermuter instance with new model and parameter dictionaries that can then be used to fit new pipeline permutations.

Version 0.9.0 - Apr 25, 2023

Changes

  • removed biopsykit.carwatch_logs module for processing CARWatch log data (moved to own carwatch package)

Bugfixes

  • biopsykit.classification.model_selection.SklearnPipelinePermuter: some bugfixes and improvements when computing additional classification metrics
  • additional bugfixes in other functions

Misc

  • code changes after switching to ruff as new linter

Version 0.8.1 - Apr 04, 2023

Added

  • biopsykit.classification.analysis: Added functions to analyze classification results
  • biopsykit.classification.model_selection.SklearnPipelinePermuter: Added methods to export classification metrics as latex tables, to merge to different SklearnPipelinePermuter instances, and to compute additional classification metrics.
  • various new util functions

Bugfixes and Improvements

  • biopsykit.classification.model_selection.SklearnPipelinePermuter: fixing random state now works as intended and produces reproducible results for real!
  • biopsykit.stats.StatsPipeline: Improved behavior
  • biopsykit.plotting: Improved layout
  • minor bugfixes

Version 0.7.1 - November 30, 2022

Bugfixes

  • bp.io.biopac:wrong index name for dataframe export when setting parameter "index = None"

Version 0.7.0 - November 22, 2022

Added

  • biopsykit.io:
    • Added functions to directly load aTimeLogger files (biopsykit.io.load_atimelogger_file) and to convert time log files into a dictionary with start/end times (biopsykit.io.convert_time_log_dict)
    • Added support for loading data acquired from the Biopac system (biopsykit.io.biopac)
  • biopsykit.plotting: Added function to conveniently plot feature pairs (biopsykit.plotting.feature_pairplot), which is a wrapper function for seaborn.pairplot

Improvements

  • Adapted code to don't show pandas future warnings anymore
  • biopsykit.plotting.hr_ensemble_plot: Added "is_relative" argument to style y-axis accordingly
  • biopsykit.stats.StatsPipeline: Improved latex export for different effect sizes
  • biopsykit.io.nilspod: Improved handling of loading NilsPods files
  • biopsykit.utils.data_processing: Added support for data dicts with tuples instead of strings as keys

Bugfixes

  • some small bugfixes

Version 0.6.1 - August 19, 2022

Improvements

  • biopsykit.signals.ecg.EcgProcessor: improved handling of errors occurring during ECG processing
  • biopsykit.plotting: Improved boxplot behavior due to the use of the new light palettes of fau-colors. See the StatsPipeline & Plotting Example for further information.

Bugfixes:

  • biopsykit.signals.imu.ActivityCounts: added timezone support (PR 31)
  • biopsykit.signals.imu.RestPeriods: fixed bug in computing resting periods (old implementation could lead to classifying a long period of activity as the main rest period.) (PR 32)
  • biopsykit.signals.imu.ActivityCounts: fixed bug(s) in computing activity counts (PR 33)
  • biopsykit.stats.StatsPipeline: fixed bugs in LaTeX export of result tables

Version 0.5.1 - May 24, 2022

New Features

  • biopsykit.classification: SklearnPipelinePermuter now supports regression

Bugfixes

  • further small bugfixes

Version 0.5.0 - April 21, 2022

New Features

  • biopsykit.stats: Added new statistic features:
    • biopsykit.stats.multicoll: Functions to handle multicollinearity in data
    • biopsykit.stats.regression: Functions for performing more "complex" regression analysis (currently: stepwise backward multiple linear regression)
  • biopsykit.classification: SklearnPipelinePermuter instances can now be saved as pickle files and imported again

Bugfixes

_ biopsykit.utils.time: Renamed time_to_datetime to time_to_timedelta (wrong function name)

Version 0.4.2 - March 18, 2022

Bugfixes

  • biopsykit.stats.StatsPipeline: Fixed bug of failing to generate significance brackets from within-subjects stats data
  • biopsykit.protocols.plotting.saliva_plot: Fixed wrong plot styling

New Features

  • biopsykit.plotting.feature_boxplot: Added legend_title parameter
  • biopsykit.classification.model_selection.nested_cv: Train and test indices are now saved for each fold of the nested cv
  • biopsykit.classification.model_selection.SklearnPipelinePermuter:
    • Updated "scoring" and "refit" behavior
    • Added confusion matrix, true labels, predicted labels, train and test indices per fold and flattened

Version 0.4.1 - January 24, 2022

Bugfixes

  • biopsykit.signals.imu.static_moment_detection: Now returning an empty dataframe with correct column names when detecting no static moments
  • biopsykit.classification.model_selection.SkleanPipelinePermuter: Pipelines are now correctly sorted by their mean test scores
  • biopsykit.protocols.plotting.saliva_plot: Fixed bug when attempting to plot saliva data without sample times _ biopsykit.plotting.feature_boxplots: Updated plotting color configuration: If an alpha value smaller than 1.0 is passed, the saturation is by default set to 1.0 to still have a nice color shading despite the transparency
  • biopsykit.protocols.TSST: Changed default TSST test time from 20 to 15 min

Version 0.4.0 - January 05, 2022

New Features

  • biopsykit.stats.StatsPipeline:
    • added new functions StatsPipeline.results_to_latex_table and StatsPipeline.stats_to_latex to convert statistical analysis results to LaTeX tables and to inline LaTeX code, respectively.
    • changed argument stats_type to stats_effect_type.
      WARNING: This is a breaking change! The use of stats_type is deprecated and will be removed in 0.5.0!
  • biopsykit.colors: Removed biopsykit.colors module and replaced it with colors from the new fau_colors package.
    WARNING: This is a breaking change!

Bugfixes

  • biopsykit.signals.imu.static_moment_detection: end indices of static moments are inclusive instead of exclusive

Misc

  • biopsykit.utils.array_handling: if overlap_percent > 1 it is converted to a fraction [0.0, 1.0]

Version 0.3.7 - December 03, 2021

Misc

  • Updated dependencies: now using scikit-learn >=1.0

Version 0.3.6 - December 01, 2021

New Features

  • biopsykit.io.ecg: Added functions to write a dictionary of dataframes to a series of csv files (and vice versa).
  • biopsykit.classification.model_selection: Model selection now also supports to perform randomized-search instead of grid-search for hyperparameter optimization of machine learning pipelines

Bugfixes

  • biopsykit.questionnaires.psqi: Fixed bug in computing Pittsburgh Sleep Quality Index (PSQI)

Version 0.3.5 – November 05, 2021

Bugfixes

  • further bugfixes (incorrect handling of duplicate index values) for Sleep Analyzer import data using biopsykit.io.sleep_analyzer.load_withings_sleep_analyzer_raw_file()

Version 0.3.4 – November 04, 2021

New Features

  • added function biopsykit.utils.time.extract_time_from_filename() that can parse time information from a filename
  • added function biopsykit.metadata.gender_counts()

Bugfixes

  • fixed bug when importing Sleep Analyzer data using biopsykit.io.sleep_analyzer.load_withings_sleep_analyzer_raw_file()
  • biopsykit.io.load_time_log(): time log dataframes imported using this function will now have "phase" as column name

Misc

Version 0.3.2 – October 12, 2021

Version 0.3.1 – October 5, 2021

  • biopsykit.utils.file_handling.get_subject_dirs(): replaced glob-style matching of regex strings by re-style matching. This allows more flexibility when filtering for directories that match a specific pattern.
  • biopsykit.questionnaires:
    • fixed wrong implementation of ASQ (biopsykit.questionnaires.asq)
    • added modified version of the ASQ (ASQ_MOD, biopsykit.questionnaires.asq_mod) according to (Kramer et al., 2019)
  • implemented changes as result of JOSS submission, including:
    • improved handling of example data
    • improved interface for multi-test comparison in StatsPipeline
    • improved colormap handling
    • new function to load dataframes in long-format (biopsykit.io.load_long_format_csv)

Version 0.3.0 – September 1, 2021

  • initial commit