GSoC 2022 Project Discussion

Development
21

Can I make a draft PR for MvNormal distribution based on what I’ve learned about Aesara from its documentation?

22

Yes. I’ve already started porting MvNormal few months back, but left it in the middle. Here is the PR: #5424. Feel free to continue the work either by building on top of #5424 or creating a new PR.

Thanks

23

Thank you, @Sayam753, I saw your PR #5424. Do we need to address aesara-devs/aesara#791 first to build Aesara for batched matrix inversion?

24

I don’t think aesara-devs/aesara#791 relates to batched matrix inversion. Rather it is about generalizing linalg.cholesky.

25

Thanks for the clarification; I think I’ve almost grasped the concept of batched matrix inversion and will submit a PR soon.

26

@Sayam753 from your PR #5424 I figured out which Aesara Op we need to implement for the MvNormal Distribution, but I’m having trouble figuring out the remaining Ops for other distributions.

I couldn’t understand the Aesara code at first, so I learned the basics of Aesara from its Tutorial and practiced all of the programs mentioned in it, and now I’m completely comfortable with Aesara.

But I need some assistance in collecting Ops for these distributions; is it possible to discuss this in a Zoom session?

27

Have you gone through the current implementations of multivariate distributions in PyMC?

28

I’ve managed to figure out MvNormal :disappointed_relieved:, but I haven’t started on the other distributions yet. do you have any suggestions for me?

29

Hi @Sayam753. Thank you so much for taking your valuable time to explain the project implementation over Google Meet. I’ll follow the steps and begin working on the MvNormal distribution.

I updated the proposal in response to your previous comments; please take a look and please give me some suggestions at your convenience.
Link to the proposal

Thank you very much

CC: @ricardoV94, @lucianopaz

30

Thanks for sharing the updated proposal. I’ll try reviewing it by the weekend.

Best, Sayam

1 Like
31

Thank you very much. @Sayam753 I’ll be looking forward to hearing from you.

32

Hello, @Sayam753!

My proposal has been submitted to the GSoC portal, which can be found at this link. I’ll revise it based on your suggestions. I’m looking forward to hearing from you!

Proposal link
Thank you very much.

33

Hi Purna.

Thanks for sharing the proposal. Here are my two comments:

  • In the proposed timeline, there is a subtask Generalize linalg.cholesky beyond 2D arrays in week 11-12. This should be worked out in phase 1.
  • FYI, NumFOCUS recommends writing bi-weekly blog posts during GSoC coding period. So, I think you should mention dedicating some time for writing blog posts to share your learnings/experience.
34

Thank you so much for taking the time to look over my proposal.
I have revised my proposal in response to your suggestions, which can be found here.
I’d be grateful if you could provide some additional feedback on it whenever it’s convenient for you.

Thank you very much.

35

Hi @Sayam753, it’s less than a week until pencils down. Could you please take a look at the proposal before final submission?

Thank you.

1 Like
36

image

I’m overjoyed to share that my proposal for GSoC-2022 has been accepted with PyMC under the NumFOCUS umbrella.

I’d want to express my gratitude to all the PyMC members for their assistance and support. especially @OriolAbril, @Sayam753 @ricardoV94.

I’m delighted to learn a lot from you and to contribute to the PyMC library under the guidance of @Sayam753, @ricardoV94, and @lucianopaz!

Thank you once again for putting your faith in me and selecting me for this project : )

5 Likes