8000 Add Diffeo abstraction by luisfpereira · Pull Request #1863 · geomstats/geomstats · GitHub
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content

Add Diffeo abstraction #1863

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
wants to merge 1 commit into from
Closed

Conversation

luisfpereira
Copy link
Collaborator
@luisfpereira luisfpereira commented Jun 1, 2023

This PR brings the notion of Diffeo. This intends to simplify PullbackDiffeoMetric, as well as make it more logically sound (both mathematically and computationally).

In the current implementation everything that is related with the diffeomorphism (meaning moving points and tangent vectors from one space to the other, in both directions) is implemented within PullbackDiffeoMetric. I argue these operations can live by themselves and are independent of PullbackDiffeoMetric (this PR moves them to a new class called Diffeo). On the other hand, PullbackDiffeoMetric needs access to this object.

I've changed the code in UnivariateNormalMetric (may interest you @alebrigant and @tramy1258), and ElasticMetric and SRVMetric (may interest you @alebrigant) to be according to my proposal.

This also simplifies testing, as PullbackDiffeoMetric only defines methods that we clearly associate to a metric (exp, log, dist, etc), and nothing related with point transformations.

I think this also interests you @ymontmarin.

This is just a prototype (I haven't changed the tests yet, so their failing). Looking forward for your feedback.

@luisfpereira luisfpereira added refactoring code refactoring geometry geometry related content labels Jul 11, 2023
@luisfpereira
Copy link
Collaborator Author
luisfpereira commented Jul 12, 2023

@ninamiolane, I would like to move on with this idea as we are bringing in several PullbackDiffeoMetric for full rank correlation matrices and spd matrices that would benefit from this design. Do you think the design is proper? Can I move on and fully implement it everywhere it is required?

@ymontmarin
Copy link
Collaborator

@luisfpereira I just looked, and the design seems coherent with our last discussions on this subject. I am on holiday, and I will be back on the 1st of August and will be available for review if you need me. I did not see your previous invitation; I think it has expired.

@luisfpereira luisfpereira requested a review from ymontmarin July 28, 2023 08:17
):
self._check_ambient_manifold(space.ambient_manifold)

diffeo = FTransform(
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The reading is simple, it is great

@luisfpereira
Copy link
Collaborator Author

Closing as new version as been developed more carefully and thoroughly in #1904.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
geometry geometry related content refactoring code refactoring
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants
0