LABS
by
Jay Qi
Our team is excited to share a new feature that you will see in DrivenData competitions: Community Code!
The Community Code section is a place where participants can share helpful code related to the competition. This could be anything from short snippets demonstrating some data processing to longer tutorials or analyses.
Community Code will be enabled on a competition-by-competition basis. The first competition that has this enabled is our new Unsupervised Wisdom competition to extract insights from data about older adult fall-related injuries. You can find the Community Code section here. To encourage contributions, this competition includes a $2,500 "Most helpful shared code" bonus prize to a participant selected by judges.
In the rest of this post, we'll give a quick tour of Community Code, discuss why we created it, and share some ideas for where it might go in the future.
A quick tour of Community Code¶
The Community Code section, if enabled for a competition, will be navigable from the left-hand navigation sidebar.
The main page of the section shows you a list of all posts that have been contributed, including some that have been made by DrivenData staff to help everyone get started. You can use the buttons at the top of the list to sort by recency or by number of upvotes. Click on the title of a post to view the post itself.
When viewing a post, you'll be able to see a nicely rendered view of the content. We currently support rendering Jupyter notebooks, markdown documents, and scripts with syntax highlighting.
On the right-hand side, you'll see some additional information and options about the post. You'll be able to download the original file (e.g., the .ipynb file for a Jupyter notebook), and, if the contributor included one, an environment requirements file (e.g., a requirements.txt). If you are viewing your own post, you'll also see options for editing or deleting your post. Finally, contributors can also indicate whether their post was inspired by another post. Links to inspired-by posts will be listed, and a link back to inspired posts will also be shown on the inspired-by post.
If you like a post or found it to be helpful, use the "Upvote" button. We track the total number of upvotes received by all posts, and this can be a helpful signal on which posts have gotten positive feedback from the community overall.
When you're ready to contribute some code yourself, click on the "Contribute a post" button on the main page. We look forward to seeing what you decide to share!
Our goals¶
The main goal of Community Code is to give participants a place to share helpful code with each other. Helpful code can improve everyone's collective experience participating in the competition and lead to better final submissions.
We also believe that shared code will be useful learning resource, both during and after the competition. Community Code posts are publicly available and will remain so after the competition ends. We expect that similar projects will be able to benefit from the posts of past competitions that share related datasets or tasks.
Finally, we hope that Community Code will be a convenient way for you to showcase some cool work as part of your online data science portfolio. Your post will be presented alongside context for how it addresses a real-world problem.
What's next?¶
The Community Code section is a brand new part of the DrivenData competition experience, and we're excited to follow how it gets used. We have a lot of ideas about expanding it in the future. Some of the ideas we're considering include:
- A way to browse and search Community Code posts globally across all competitions
- The ability to showcase your Community Code posts on your user profile
We'd love to hear any feedback, ideas, or questions that you may have about Community Code! Please post them to this forum thread.
As mentioned before, you can check out a live Community Code section for the Unsupervised Wisdom competition. DrivenData staff have contributed a few starting posts, such as this one about using the falcon-7b large language model. We hope that you'll decide to contribute some of your code as well. Happy sharing!