This repo is based on Awesome reading group. The idea is to organize short 45-60 minute meetings one time a week. No one is expected prepare slides and/or read the papers in depth. This makes it agile and highly sustainable.
This goes against the initial idea of the *Awesome reading group* of having shorter meetings multiple times a week, but in the context of Kalido I feel that would just not work (at least not at this point in time). The changes we've thought of to adjust for this: weekly session, longer session, group multiple papers together with a common theme in one session while having a living backlog of papers.
Please feel free to give any feedback on current organisation as a new issue to this repository - it can be adapted to our needs.
The time/frequency can be adjusted based on your feedback. The objective is to share and discuss interesting papers together on a regular basis.
- ➕ 👍 Add papers and vote on papers you would like to read/discuss. Tag your papers with topics.
- 🌝 🌚 Everybody is free to attend or not attend;
- 📝 Give the papers a read before as much in details, focus on specific parts or choose a specific paper from that theme
- 📢 At the beginning of the sessions a volunteer does a brief introduction of the theme / topic
- 💬 Attendance (usually groups 3-8 participants) can discuss, share impressions, ask questions
- ⏳ Try to keep it under 60 minutes.
- ✏️ take 5 mins to write your reflections on the discussion or add extra thoughts.
1 - ADD and VOTE on papers
To add a paper, create a new issue for it and tag it with appropriate topics. Feel free to add new tags if what the paper tackles does not exist in our collection of tags. For instance, if adding the original BERT paper you should add as tags: NLP, Transformers, Unsupervised Learning, Self-Supervised Learning.
You may only put the title and the link to the paper, or add any comments to help people to understand why it is worth reading.
2 - The topic with the most votes each week gets selected
A github discussion is started with the selection so that we can better organize who takes which paper and when to schedule the reading group session. Everybody goes through papers, selects the titles they are interested it, try to read through to get an idea of the paper. Then you submit your reading preference in a comment to that discussion. You may add a short comment about 'why it is interesting, worth reading, what it is about' or any reminder for yourself.
3 - Read the selected paper according to one's time/knowledge
- You don't have to read everything
- You are not supposed to understand everything
- The purpose of the multi-view discussions is to save time reading all papers in details as different participants will have focused on different parts of the paper.
4 - Attend the reading group
- The main value of such reading groups is in the discussion; even if you didn't read the paper in depth - everybody pays attention on different things, which are shared during the discussion. Thus, it allows you to cover each paper more than if you would have read it by yourself.
5 - After the session add your thoughts as a comment on the github discussion