Would there be any interest in a Covariance module?

[zdk123]

Similar to the scikit-learn module. I'm primarily interested in implementing the Graphical Lasso, but empirical and shrinkage covariance estimators as well.

[YuhanLiin]

I think you can put that into a new algorithm crate. What do you think @quietlychris @bytesnake

[bytesnake]

Hey Zachari, a covariance estimation crate would be a great addition to Linfa! Sparse covariance estimation should be not too hard, you can probably reuse routines for elasticnet solvers. Question is how to represent a structured covariance, ndarray currently only supports dense, no banded or sparse representation. If you have questions please let us know 26.02.2023 13:28:01 Yuhan Lin ***@***.***>:

…

I think you can put that into a new algorithm crate. What do you think @quietlychris[https://github.com/quietlychris] @bytesnake[https://github.com/bytesnake] — Reply to this email directly, view it on GitHub[#293 (comment)], or unsubscribe[https://github.com/notifications/unsubscribe-auth/AAHRRKKYQZ4QIQ6ZMQNK543WZO4EXANCNFSM6AAAAAAVITKABI]. You are receiving this because you were mentioned.[Verfolgungsbild][https://github.com/notifications/beacon/AAHRRKLXW77TZN5XLPAOIOTWZO4EXA5CNFSM6AAAAAAVITKABKWGG33NNVSW45C7OR4XAZNMJFZXG5LFINXW23LFNZ2KUY3PNVWWK3TUL5UWJTSWE72MM.gif]