Good first issues & getting started, for new contributors

A number of good issues to start working on as a new contributor. Contributions to documentation are especially appreciated.

getting started with contributions

1. Say hello on Slack and get set up for development, see instructions in the developer guide
2. Pick a “good first issue” to work on, see a collection in this summary issue, or from this list suggestion: pick something small with simple content to learn the “process”
3. Feel free to attend the regular community collab sessions or one of the topic specific stand-ups and tech sessions (see schedule on slack)
4. Once your first PR is merged and you’ve seen how things work, think about your regular time commitment. Optionally, continue attending the Friday community collaboration sessions and stand-ups; or, optionally, apply for mentoring

introductory and user testing

• work through the sktime tutorials and record feedback, you can post it in this issue #1447
• work through user testing sheets for forecasting or time series classification, record feedback and send us your user testing sheet with code (e.g., attach to this issue)
• open an issue with a time series related algorithm that you would like to see implemented or interfaced in sktime
• work through the installation instructions to install a development version of sktime. Record carefully things that are unclear or don’t work as described, return the full feedback to us (e.g., in an issue).

contributors new to open source

• documentation: many code files are missing module and class/function docstrings. Pick a file and start adding these. A good starting point are estimator algorithms, these should follow the general documentation style outlined in #1148 or the extension templates. If you start, post the name of the estimator you are working on in #1148; when you open a PR, refer to the issue.
• documentation: many code files are missing examples. Pick a file and start adding these. a good starting point are estimator algorithms, and time series classifiers, see #932 and #1148

NOTE: #1148 and #932 are closed, but there are still modules without good docstrings. They are just no longer tracked by the issue, which covered the priority items only. So, feel free to go through the code base and help improve docstrings!

small-to-medium documentation and technical writing tasks

• other smaller issues in the documentation milestone
• the user guide is incomplete. Help is appreciated - the best approach may be converting the tutorials in a well-written guide.
• the tutorial for time series classification should be re-factored, in the style of the forecasting tutorial
• extension templates for time series regression and time series annotation should be written. The current extension templates can be used as an extension template template.

small-to-medium python/coding tasks

• current open small tasks with a “recipe”: https://github.com/alan-turing-institute/sktime/issues/3429
• pick an unverified bug from the bugfixing board: https://github.com/alan-turing-institute/sktime/projects/18 and try to reproduce it. Post operating system, python version and (if relevant) package versions, and whether you can yes/no. Investigate further if the bug looks simple - but note that bugs can be anywhere between easy to very difficult to track down.
• Pick an algorithm to implement or interface, according to the extension templates. Possible choices are forecasters (univariate or multivariate), time series classifiers, time series regressors, annotators, distances, kernels - these can be existing feature requests, or new suggestions. We currently do not recommend transformers as they undergo refactoring.
• implement get_fitted_params functionality in estimators where this is missing (currently most of them). Related issue for composite estimators: https://github.com/alan-turing-institute/sktime/issues/1497

mid to longer term tasks

For contributing across a period of weeks or months, consider joining one of the major workstreams and weekly stand-ups on Fridays. Join the slack (see link in README) and say hello on the contributors channel.

Current workstreams, overview here https://github.com/alan-turing-institute/sktime/projects?type=classic:

• forecasting https://github.com/alan-turing-institute/sktime/projects/3 and https://github.com/alan-turing-institute/sktime/projects/25
• deep learning, benchmarking https://github.com/alan-turing-institute/sktime/projects/31
• time series annotation, segmentation, change points https://github.com/alan-turing-institute/sktime/projects/27
• pipelines, pywatts integration https://github.com/alan-turing-institute/sktime/projects/30
• docs, maintenance, testing https://github.com/alan-turing-institute/sktime/projects/26

challenging tasks

✨ Your time to shine! ✨ Talk to a core dev before starting with these. Also have a look at the roadmap #228

• deep learning interfaces and models, see workstream above
• get_report interface for model inspection #971
• refactor/redesign: forecasting interface for stream update/prediction #982
• help build/design new experimental module: distances/kernels
• help build/design new experimental module: annotations (segmentation, change point detection, outlier detection)
• help build/design new experimental module: sequence alignment

[aiwalter]

Hi @gepitis , you could open a new issue with ideas of improvement that you have in mind

[fkiraly]

Where can I send my feedback on the tutorials?

I just opened an issue here https://github.com/alan-turing-institute/sktime/issues/1447 thanks for any feedback!