Hugo Delatte

Yes also a good idea. It's related to https://github.com/skfolio/skfolio/discussions/41#discussioncomment-8889409 I'll list all arguments and features that I have in mind taking inspiration from that previous discussion and the polar implementation...

Hello, the solution would be to implement Transformers. It's in our roadmap. For more details, it's related to [this discussion](https://github.com/skfolio/skfolio/discussions/60).

Thank you for proposing to implement it. The work has already started: I'll push it in a feature branch soon so you can participate in its development. I'm referring to...

Yes if you fill with a constant value, you get zero variance. In that case, that asset should be removed before the optimization instead. We could set it as a...

It's because of either bug or new features from the dependencies. For example, plotly had a bug in the histogram plot so we had to bump to the version that...

Yes agreed, good idea. Do you wish to create a PR for it otherwise I can do it? The implementation should be relatively similar to `plot_cumulative_returns` I think.

Implementation done in dev branch here: https://github.com/skfolio/skfolio/blob/128f619960494f827fb2205e1a9d5cca383092d8/src/skfolio/population/_population.py#L455

Do you have more details or literature references in mind? All precision matrices from scikit-learn covariance estimators are computed using the same function: `scipy.linalg.pinvh`. Also, it has been shown that...

Yes we can include an example. The `GraphicalLassoCV` iteratively compute the precision matrix and the covariance and save them into its class attribute `covariance_` and `precision_`. I just doubled check...

We don't use directly the precision matrix because the problem is solved using Second-order cone programming which needs the Cholesky decomposition. For examples, have a look at https://docs.mosek.com/portfolio-cookbook/markowitz.html#conic-formulation. Also, SOCP...