Getting started

Here are some ideas and notes on what we should do to get started. Feel free to edit/add to them.

Project management

Project management on GitHub: alan-turning-institute/MLJtime.jl (milestones, API proposal, issues, code reviews)
Chat on slack, feel welcome to ask any questions, I'm online most of the days
Technical discussions on GitHub ("communication is documentation"),
Blog posts will be hosted over here.
Student is available throughout the summer except 15 days due to exams (exact times tbc due to lockdown)
In the summer, we'll have the chance to give feedback via the student and mentor evaluations.
Set up weekly meeting: Mon, 2pm UK time

Agree on rough git workflow (branches, commit conventions, PRs, etc), see this website for an overview
Share files in Google Drive folder "GSOC2020-Time"
Set up template for weekly reports (should include: summary of weekly deliverables, link to GitHub milestone, whether milestone has been completed, some reflection on what worked well, what didn't and how much time you spent for different tasks): weekly reports
We agree on Blue code style, make sure you can easily ensure that you follow it (IDE settings)

@aadesh, I know you've opened the PR on the MLJ repo, have you also introduced yourself on the MLJ and Julia slack channel? It's important to connect with other GSoC students and the community and see if any other people are interested in machine learning with time series, they may have good ideas on where to start, see e.g. this issue that was posted on sktime, feel free to comment on it and link it to our new repo.
Accept Code of Conduct (CoC) and add it to the repo, not sure if MLJ has one, otherwise I propose to follow sktime's CoC, please adapt where necessary, there's also the GSoC CoC
MLJ & Data channel on Julia slack will be great to discuss challenges.

Please set them up on GitHub as milestones, so that we can edit/update them there
agree on time frame for completing them

agree on framework to start with (forecasting, classification, etc): classification
identify challenges (e.g. for classification, choice data container; for forecasting, cross-validation, etc)
agree on whether to wrap sktime or re-implement algorithms
write API design proposal, including overall introduction/motivation, discussion of key features, limitations and alternative solutions, list of classes/methods to implement, for an example, see sktime's forecasting API proposal
sktime is currently undergoing some major refactoring, please refer to the dev branch and this PR for recent changes

implement one algorithm implementation in Julia (e.g. the time series forest classifier), and a wrapper for sktime classifiers
run benchmark on time series classification repository, compare against sktime and tsml results