Beta 0.5.0rc2

What's Changed

• Move tensors to correct device by @karl-richter in #1010
• Uncertainty: Conformal Prediction V1 by @Kevin-Chen0 in #802
• Add a main script to print neuralprophet version by @noxan in #974
• Follow-up tensors on wrong device by @karl-richter in #1023

Deprecations

• Change plotting_backend deprecation warning for implicit and explicit matplotlib use by @LeonieFreisinger in #1006

Bug fixes

• Fix string Interpolation by @poweroftrue in #1026

Refactors

• Meta name bool by @alfonsogarciadecorral in #997

Internals and website

• [docs] New tutorials section for docs by @noxan in #976
• [dev-ops] Type annotation validation GitHub action by @noxan in #902
• Add pull request template for Github by @karl-richter in #915
• Fix typings and constructor super call for pinnball loss by @noxan in #1019
• [website] Fix typo in preview of website description by @noxan in #1012
• Fix label links in contributing guidelines by @noxan in #1024

New Contributors

• @poweroftrue made their first contribution in #1026

Full Changelog: 0.5.0rc1...0.5.0rc2

Beta 0.5.0rc1

Relevant new features:

• #961 GPU (and other accelerator) support
• #888 Plot multiple sets of parameters due to Glocal model as mean and 10/90 percentiles.
• #779 New panel in plot_forecast depicting uncertainty
• #884 Plot select panels in plot_parameters and plot_components

API and modeling changes:

• #837 Migrate backend to PyTorch Lightning
• #929 remove support for dictionaries as inputs. Please use a single DataFrame with an ID column for each time series
• #984 deprecation warning: upcoming change of plotting backend to plotly (matplotlib support will be dropped in future)
• #927 remove addition of residuals to forecast dataframe and subsequent plotting

Important fixes and enhancements:

• #978 Fix None dtypes to NAN
• #968 Sort changepoints
• #919 backwards compatibility for plot_latest_forecast
• #981 Add type annotations for NeuralProphet class
• #853 new argument for setting custom quantile in plot_parameters
• #808 Do not plot weekend seasonality for business day data frequency

Additionally, many smaller bugfixes and improvements to the codebase were also introduced.
For details, please view the merged Pull Requests.

Beta 0.4.2

Relevant new features:

• #916 Introduce "Glocal": global-local seasonality and trend
• #733 Introduce TorchProphet: Making NeuralProphet compatible with Facebook Prophet code, with tutorial
• #831 Tutorial: Migration from Prophet to NeuralProphet
• #779 Introduce uncertainty panel in forecast components plot

API and modeling changes:

• #829 API change: Rename "yhat" to "origin" in get_latest_forecast() and plot_latest_forecast()
• #909 Remove regressor if training set has one unique value (instead of throwing error)

Important fixes and enhancements:

• #853 Support setting custom quantile in plot_parameters()
• #865 Allow component plot with locally-normalized global df
• #920 Restructure documentation website
• #791 Fix: re-scaling of multiplicative components in forecast df

Additionally, many smaller bugfixes and improvements to the codebase were also introduced.
For details, please view the merged Pull Requests.

Beta 0.4.1

Relevant changes and new features (excluding documentation improvements and bugfixes):

• #603 introduce __version__
• #619 Allow missing values left unimputed - drop affected areas
• #641 Global modeling: accept DataFrame with IDs (deprecating dict of df)
• #658 introduce Future regressor positivity constraint
• #664 increase torch minimum required to 1.8.0
• #669 introduce uncertainty estimation via Quantile Regression
• #671 introduce Plotly support
• #673 add Network Visualization tutorial
• #691 introduce save and load
• #701 introduce get_latest_forecast
• #714 support regularization for lagged regressors
• #800 increase last trend segment stability, setting changepoints_range to 0.8

Additionally, many bugfixes and improvements to the codebase and documentation were also introduced.
For details, please view the merged Pull Requests.

Beta 0.3.2

Included in this release:

• bugfix for Torch 1.9.0 (missing torch.pi)
• New docstrings in Numpy format for most files
• increase training time for better stability
• speed up learning rate range test
• updated tutorial notebooks
• add benchmarking test coverage
• bugfixes to benchmarking framework
• ! API change: AR regularization: move from ar_sparsity to ar_reg
• documentation for Global Modeling

Beta 0.3.1

Included in this release:

• Now supporting use of multiple time-series datasets to train a single model (global modelling)
• example notebook for global modelling
• data frequency argument optional - automatic detection
• improved documentation and docstrings
• support for local/individual normalization of time-series when working with multiple datasets
• introducing forgetfulness: Skew model fit towards more recent observations
• widen range of default number of epochs
• add notebook for use of live-plot-loss
• improve docstrings to show up in sphinx (API documentation)
• bugfixes

Beta 0.3.0 (some API changes)

• Add benchmark framework
• Support panel datasets with global modelling
• Add minimal verbosity option to fit method
• Allow no metrics
• Repeat learning-rate range test 3 times, use log10 avg
• Update energy example notebook
• Require passing dataframe for validation data while training
• Update how to build documentation added to Contributing
• Documentation using sphinx (before: mkdocs)
• Now optional: using make_future_dataframe
• avoid double calls to normalization and fill missing data methods
• New notebook guiding how to collect predictions
• Make raw predictions available to user
• Embed Tutorials in documentation page
• Embed Docstrings in documentation page
• move data to ourownstory/neuralprophet-data repository
• New energy notebook on ERCOT data
• Support more types of custom loss functions
• remove reliance on attrdict, use dataclasses instead
• improved plotting legend
• fix issues

Beta 0.2.8 (many improvements)

• Robustify automatic batch_size and epochs selection
• Robustify automatic learning_rate selection based on lr-range-test
• Improve train optimizer and scheduler
• soft-start regularization in last third of training
• Improve reqularization function for all components
• allow custom optimizer and loss_func
• support python 3.6.9 for colab
• Crossvalidation utility
• Chinese documentation
• support callable loss
• Robustify changepoints data format
• require log_level in logger util
• Rename tqdm, remove overbleed option
• Reg schedule: increasing regularization in last third of training
• bug fix in plot country holidays
• Add Energy datasets and example notebook
• disable log file by default
• add double crossvalidation
• improve tests
• Buxfixes

Beta 0.2.7 (bugfixes, tutorial notebooks)

• example notebooks: Sub-daily data, Autoregresseion
• bugfixes: lambda_delay, train_speed

Beta 0.2.6 (auto batch, epochs)

• Auto-set batch_size and epochs
• add train_speed setting
• add set_random_seed util
• continued removal of AttrDict uses
• bugfix to index issue in make_future_dataframe