Skip to content
This repository has been archived by the owner on Jun 22, 2024. It is now read-only.

Latest commit

 

History

History
158 lines (128 loc) · 10.7 KB

CHANGELOG.md

File metadata and controls

158 lines (128 loc) · 10.7 KB
Raw

0.3.1 - Unreleased

Added

  • Add alpha to rolling mean (#225)
  • RiverDriftDetectionCondition for wrapping the river drift detection algorithms (#221)
  • Add a module for merging shifted time series (#2179)
  • Add a module for statistic feature extraction (#217)
  • Add a tsne summary module (#229)
  • Add a module for inserting synthetic concept drift (#226)
  • Add a TSTR Summary (#232)
  • Add a confusion matrix summary (#236)
  • Add a module to insert unusual behaviour into a time series (#240)
  • Add F1Score Summary (#234)
  • Add a discriminative score summary (#238)
  • Add NPY Callback (#247)
  • Add Select transformer (#257)

Changed

  • Make the inputs of the PNN more flexible (#228)

Fixed

  • Enabled installation for macOS(#241)
  • Adapt CD Condition to new river api (#249)
  • Wrong sine cosine encoding in calendar extraction module (#36)

Deprecated

  • Sampler and ClockShift. Will be removed in version 0.5.0 (#257)

0.3.0 - 2022-31-03

Added

  • Add average module for pre-processing (#166)
  • Metric can be calculated on cutouts (#149)
  • Add custom scaler module for pre-processing (#167)
  • Add a MASE Summary (#148)
  • Add ensemble module for post-processing (#165)
  • Online Start Feature (#189)
  • Enable Multiple RefitConditions (#192)
  • Periodic Refit Condition (#192)
  • Add AnomalyGeneration modules (#195)
  • Add get_min_data method in modules to enable them to specify how much data they need. (#147)

Changed

  • Retraining is triggered after all steps are transformed (#117)
  • The sklean-wrapper includes now methods based on the SelectorMixin (#185)

Fixed

  • Try to handle non matching shapes in metrics. If not possible raise an exception. (#154)
  • Fixed Sampler-module, samples now in forwarding direction (#174)
  • Fixed wrong check of if summary should be returned in _run method of pipeline (#157)
  • Fixed Pipeline.to_folder() for modules that store parameters in numpy format by adding a special JSON-encoder for types that are not json-serializable by default (#175)
  • Added missing summaries to pywatts.summaries.init.py (#183)
  • Naming in the Metric Summaries (#193)
  • Fixed wrong time dimension name in ComputationMode.Transform (#199)
  • Fixed online execution of subpipeline (#203)
  • Enable to store a Keras Model with custom objects as custom loss functions (#17)
  • Fixed online execution of stacked pipelines (#213)

Deprecated

Removed

  • MAECalculator and RMSECalculator.

0.2.0 - 2021-30-09

Added

  • Logger for modules is defined in Base (#77)
  • Add parameters to RMSECalculator so that it can caluclate a sliding rmse too. (#23)
  • StatsmodelTimeSeriesModelWrapper. (#29)
  • Optional pipeline path and less aggressive FileManager directory creation (#94)
  • Add fit_method parameter to FunctionModule (#93)
  • SummaryFunctionality (#34)
    • BaseSummary Module, from which summary module should inherit.
    • SummaryStep, which handles the execution of the summary module.
    • RMSESummary and RollingRMSE as consequence of the usage of summary modules.
    • Time needed by the fit method is recorded by the summary. (Part of the summary.md)
    • Section about how to get results of a pyWATTS pipeline.
    • Enable more generic usage of summaries, including different output formats (#129)
  • Add MAE as another summary. (#99)
  • Add summary parameter to train and test to ensure backward compatibility. (#127)
  • Add Min amd Max as new summaries (#105)
  • Add MAPE as new summary (#104)
  • Add a RunSetting for setting run specific settings in the steps (#150)
  • Add Slicer to slice data in a numpy like manner, i.e. a[start:end] (#152)

Changed

  • Remove parameter step.stop. Instead we call the method _should_stop on the previous steps. (#25)
  • Change metrics to summaries (#115)
  • Change indeces and indices to indexes (#102)
  • Restructure the modules folder (#114)

Deprecated

  • Usage of RMSECalculator. Will be removed in version 0.3. Calculation of metrics should be a summary and not a module. (#34)
  • Usage of MAECalculator. Will be removed in version 0.3. Calculation of metrics should be a summary and not a module. (#115)

Fixed

  • Converts the input for the keras model in the kerasWrapper from xr.Dataarray to np.array (#97)
  • Missing call of _get_rolling in Rolling Base (#76)
  • Fix save and load of keras Wrapper (#91)
  • Import of pytorchWrapper via wrapper.init
  • Fix the mask for weekends and public holidays in rolling_base (#107)
  • Improved Online Learning Performance (#18)
    • Add a cache for the most recent value, since this value is often requested by successing modules.
    • Remove unecessary copies
    • Avoid uncessary execution in should_stop
    • Improve performance in sample_module
    • Improve performance in trend_extraction
    • Callbacks are only executed if the step is finished. No intermediate results are plotted.
  • Fix the extracted time index in rolling RMSE (#124)
  • Use raise from if an exception is raise because an other is raised before for retaining the original stack trace (#123)
  • Function module crashed because of missing is_fitted = True (#144)

0.1.0 - 2021-25-03

Added

  • Added integration tests to github actions by executing the examples in root directory (#47)
  • Implementation of the profile neural network (#71)
  • Imports from the init files for exception, modules, and wrapper. (#27)
  • Add rolling_variance, rolling_kurtosis, rolling_skewness. (#28)
  • Select the kind of groupby in the rolling function by an enum (#28)

Changed

  • Remove plot, to_csv, summary. Instead we add a callback functionality. Additional, we provide some basic callbacks. (#16)
  • Change of the API (#16)
    • Use keyword arguments instead of list for determining the input of modules. keyword that start with target are only fed into the fit function.
    • Adapt the modules, so that they are compatible with the new API
    • Use xarray DataArray instead of xr.DataSets for exchaning the data
    • Remove collect step, since through the new API they are not necessary any more
    • Add ResultStep, for selecting the desired result if a moudule provides multiple ones.
    • Remove whitelister, instead columns can be selected via square brackets.
  • CalendarExtraction: Remove encoding, for each encoding and suitable feature a new feature is created. E.g. month_sine. Additionally further calendar features are added. E.g. monday, tuesday, .. and cos encodings. For the different features, a enum type is defined.

Deprecated

Fixed

  • Fixed pipeline crashing in RMSE module because of wrong time index (#39)
  • Fixed bracket operator not working for steps within the pipeline (#42)
  • Fixed dict objects could not be passed to pipeline (#43)
  • Fixed old parameter in sample_module (#67)
  • Fixed array creation in trend extraciont (#69)

[0.0.1] - 2021-MM-DD

  • Train and test methods of pipeline returns data
  • Add PipelineStep for better management of Subpipelining
  • Replace pipeline.run with pipeline.train and pipeline.test
  • Remove CSVReader and add a parameter data to pipeline.run() method. Data can either be a pandas dataframe or an xarray dataset.