Follow-up on Pull Request Submission for Manopt

[Jinghui Yuan]

Dear Nicolas Boumal,

I hope this message finds you well. I have submitted a pull request on the Manopt GitHub repository and have also sent you an email regarding the same. I am writing to confirm whether you have seen it and to express my eagerness to discuss it with you. If there are any issues with the code, I would appreciate your feedback.

Best regards,

Jinghui

[Nicolas Boumal]

Dear Jinghui -- yes, well received, thank you. This will take a bit of time to look into, but I'll be in touch.   -- Nicolas

[Jinghui Yuan]

Hello Boumal! I’m here to share some great news! After multiple rounds of checks, our paper is now live on arXiv: https://arxiv.org/abs/2503.20505. We’re thrilled that our work has already caught the attention of the machine learning community!

In the three weeks before uploading the paper, we had only open-sourced the code on GitHub (https://github.com/Yuan-Jinghui/Riemannian-Optimization-on-Relaxed-Indicator-Matrix-Manifold?tab=readme-ov-file), but even then, we received over 34 stars and several emails discussing the work.

While presenting our contributions, we also highlighted Manopt — the renowned manifold optimization toolbox. Its seamless configuration and robust features have earned widespread praise from the community.

Over these three weeks, we fixed typos in the paper, rigorously reviewed the code, and added detailed comments. If you have any questions about the paper or code, we’d love to hear your feedback! Additionally, if there are specific checks needed to better integrate with Manopt, we’re more than happy to proactively address them. Looking forward to your thoughts!

[Ronny Bergmann]

Dear Jinghui,

I am not Nicolas Boumal, so sorry for a comment here.

I noticed that you have both an own repository (congratulations on so many stars already) and did a PR on Manopt. Would that not double your code base?

I personally would avoid having to maintain two copies of the code, and deleting your current repository is also a bit sad if it is now already known to some. Note that for Manopt in Matlab also other parts are in their own repository like the two constrained algorithms ALM and EPM are available at https://github.com/losangle/Optimization-on-manifolds-with-extra-constraints.

So instead of the current PR you could also 

* keep your repository and add installation instructions in the Readme

* do a smaller PR that adds a note to the Manopt documentation where to find your manifold

Another advantage is, that a personal repository gives you a bit more visibility. Sure there is no true/false answer here, but especially with the stars and hence visibility you already have, maybe that is a good alternative as well.

Best regards,

Ronny

[Jinghui Yuan]

Hi Ronny and Nicolas!

Thanks a million for the super wise advice, Ronny! You’re totally right — keeping our own repo makes maintenance easier and gives us more flexibility. We’re happy to follow your suggestion!

Our code is 100% built on Manopt! We’ve already added a note in our README saying, “Please ensure that you download and install manopt (MATLAB version) before using this code.” 😊

Second, about adding a tiny PR to Manopt’s docs: I’d love to do that! My idea is to add a short note (like a short  RIM.md) in Manopt’s multinomial folder. Would that work? I’ll wait for your and Nicolas’ thoughts before proceeding!

Oh, and here’s some fun news! On Bilibili (a huge video platform in China, like YouTube), tutorials about Manopt and manifold optimization are getting tons of attention — one video hit 20k+ views!  For many Chinese researchers (including me!), Manopt is the gateway to this field, so It’d be an honor to contribute back!

Let me know your thoughts! Super excited to hear from you both.

Cheers,
Jinghui

[Ronny Bergmann]

Hi Jinghui,

ah, I did not check carefully enough your Readme there. That is exactly what I meant :)

For the website you would need Nicolas again, see the repository at https://github.com/NicolasBoumal/manopt-web from which manopt.org is created. I am no involved in that development, but a PR there might be a good idea if Nicolas agrees. Maybe on the about page?

I think the manifold pages are autogenerated, but that is again something Nicolas can answer better than me.

I am not involved in the homepage and only a little bit in Manopt, since I switched from Matlab to Julia quite some years back.

I am developing the Julia version (Manopt.jl for the algorithms and Manifolds.jl for a library of Riemannian manifolds); so if you ever feel your manifold might be nice to have in Julia as well, we could look into that, it could be a nice project for a student.

Best regards,

Ronny

[Jinghui Yuan]

Hi Ronny!

I love your idea! I think your suggestion is to add a new manifold description on https://www.manopt.org/manifolds.html instead of making a direct PR to the repository! That's awesome! I'm happy to do this. If Boumal supports it, I'll close the original PR in Manopt and submit a PR to this page instead — just one line of description and a link. Waiting for Boumal's reply (I’m not sure if he gets notifications from our conversation or how to @ him, haha).

Thank you for introducing me to Julia, Manopt.jl, and Manifolds.jl! I just learned that Julia is a high-performance language especially suited for optimization, with unlimited potential. Manopt.jl and Manifolds.jl are the most important manifold optimization packages in the Julia community, with huge influence! I’m truly honored by your invitation! Although I’m not familiar with Julia yet, I can definitely say I’m willing to contribute here! Though it might take me about a month to learn, I’ll work hard on it!

Looking forward to more suggestions from you and Boumal!

Best,
Jinghui

[Ronny Bergmann]

No. That page lists manifolds that are available in Manopt itself.

I specifically wrote to add it to the about page. To be precise, this list https://www.manopt.org/about.html#related-software

If you currently do your research in Matlab you should probably first stay on one language. This was not meant as an invitation or a task to directly do. Merely a rough idea for some (maybe even distant) future.

- Ronny 

[Jinghui Yuan]

Hi Ronny,

Apologies for the earlier confusion — crystal clear now! Adding our work to the Related Software section is also a great path. I will submit a PR to the manopt-web repo after getting Boumal's suggestion.

Your input is gold! If we dive into Julia after stabilizing the MATLAB version, you’ll be our first go-to expert. Let’s keep in touch!

Best,
Jinghui