GSOC Project Proposal Expectations

40 views
Skip to first unread message

Saurabh Daalia

unread,
Mar 27, 2019, 4:53:58 PM3/27/19
to Yellowbrick
Greetings Mentors,


Since Submission Period for Student Applications has begun, and I am moving ahead to draft a proposal for contributing to Yellowbrick. 
Having said that I have some doubts and questions that I hope to get inputs on.
 
Q1) Is there any specific Proposal Template that I should stick to? 
[NumFocus has given a basic template (here) though, but anything specific mentors want to be added to the proposal]

Q2) After going through Idea List of Yellowbrick, Integrating Support for Probabilistic Calibration Curve for Tree-based models is what I am more interested to contribute to. Having said that, do mentors expect the student to look through other issues which are currently opened which are listed under Milestone v.10

Q3) Since this particular project idea is about adding new visualizer (feature), apart from documenting the approach and methods that will be implemented, what other specifications and aspects that need to be explained in the proposal itself. 
For Example:
  • Should Testing be included in the project proposal
Q4) Being a former technical lead for my University's Technical Student Society, I am planning to organize a small workshop for students on topic "Machine Learning 101 with Yellowbrick"  to discuss about capabilities of Yellowbrick Library during Summers, thus would it be okay to mention this under my project proposal as well? 

Q5) What should be the proper process to get the proposal reviewed by mentors ASAP. as it is strongly advised to do before submitting the final proposal?

I am completely aware that this post is a bit lengthy, but I feel answers to these questions are very important to draft a good proposal. 

Thank you 
Saurabh Daalia

Benjamin Bengfort

unread,
Mar 27, 2019, 5:12:09 PM3/27/19
to Yellowbrick
Hello Saurabh, 

In answer to your questions:

Q1: Please use the NumFocus template - we don't have a separate template that we're using. 

Q2: End-to-end development of a visualizer or suite of visualizers is probably the best bet for GSoC and for the proposal. My recommendation is to take on 1-3 of the visualizers in the project ideas but you can also feel free to include visualizers or ideas from our issues list. Our expectation is that you will focus on the development of the visualizer(s) and will not be responsible for other issues in the milestone or on the issues page. 

Q3: As described in the contributor's guide, developing a visualizer encompasses many tasks from testing through development. For the purposes of the student application, I would recommend the following high-level outline for each visualizer:

1. Experimentation: come up with an example or two with different models/datasets to prototype the visualizer
2. Development: develop the Visualizer and quick method and justify the API choices being made (e.g. where does the Visualizer fall in the YB class hierarchy?)
3. Testing: create a suite of unit tests and image similarity tests that ensure the visualizer is robust to usage and protected from regression
4. Documentation: document the visualizer in the scikit-yb docs
5. Case studies: write a blog post on the use of the visualizer in real life
6. User feedback: get feedback from users and refine the visualizer

You can broadly sketch out your approach and methods to meet these steps. 

Q4: Yes, absolutely mention the workshop! 

Q5: Ping us when you have the proposal ready and we'll get someone to take a look at your proposal. 

Best Regards,
Benjamin Bengfort

Saurabh Daalia

unread,
Mar 27, 2019, 5:27:55 PM3/27/19
to Yellowbrick
Hello Benjamin,

Thank you so much for such a detailed answer, this cleared all my doubts and queries regarding the proposal.
Really appreciate your answer. Looking forward to completing the initial proposal and getting it reviewed. 

Thanks and Regards
Saurabh Daalia

Saurabh Daalia

unread,
Mar 30, 2019, 12:26:16 PM3/30/19
to Yellowbrick

I was hoping to get input about Title of the Proposal, in case we are mentioning about more than 1 Visualizer in the proposal, should we add all the visualizer in the title(which makes it lengthy) or should we mention only 1 based on higher priority, as I am not sure if it is mentioned in Proposal guidelines to write about two different projects in proposal

Would it make sense to give a broad title such as "Increasing Scope of Classification Visualization and Clustering Visualization"   or to give title with Specific Visualizers ?

Thank you and Regards
Saurabh Daalia 

aem...@georgetown.edu

unread,
Mar 30, 2019, 1:43:13 PM3/30/19
to Yellowbrick
Hi Saurabh,
Thanks for your message.  We trust you to make that choice.  We are not "overly concerned" with the length or lack of length (although don't write a novel!) but do be sure to focus on what you think is most important in terms of what you believe will move YB forward.  Clarity and conciseness counts for a lot as we're all busy professionals and have other obligations (same as yourself).  I hope that helps.
Thanks,
Adam
Reply all
Reply to author
Forward
0 new messages