Michael Baudin

@jschueller : I think that this PR is good to go.

That is almost ok, but I think that the users of POD update the decomposition when a new observation is available. But there is no update feature up to my...

We should wait, I think, for @adutfoy's input on the meaning of the equations.

@jschueller : I think this is ready.

Thank your @adutfoy !

I experience the same issue in a particular case. I write below the smallest case which fails. In this case, the PCE only has a single coefficient. Using a non-sparse...

@regislebrun : Any idea?

@jschueller : The new `LinearModelResult.getDesign` method will allow to implement and experiment with other methods to create sparse PCE. Meanwhile, we cannot use OT and have to use Numpy. So...

@jschueller , @regislebrun : please review this PR.

@jschueller, @josephmure : Can you please review this PR for a final round?