Tutorial

[paraynaud]

trigger PR, begin of the first pass

[paraynaud]

The reference is the first article talking about partial separability. There is no copy available on the net. Do I change it for another reference with a DOI?

I have to let

U1 = [1, 2]
U2 = [2, 3]

goes trhough, because it must be evaluated to be used in the following functions. I don't know to how make it quiet :/

[dpo]

The reference is the first article talking about partial separability. There is no copy available on the net. Do I change it for another reference with a DOI?

No, I was wondering if the reference was correct. It looks odd. In our paper, we have

A. Griewank and Ph. L. Toint. Partitioned variable metric updates for large structured optimization problems. Numer. Math., 39:119–137, 1982.

Ok for U1 and U2.

[paraynaud]

I don't know if there is an advantage for a PBFGS structure, except if this is an alias type of Elemental_partitioned_matrix that eases the update function. The flexibility of Elemental_partitioned_matrix allow to vary the partitioned update applied on the partitioned matrix.

About the code changes :

• Make PBFGS, PSR1, PSE structure may take time.
• Rename some methods to define PBFGS, PLBFGS, PLSR1... partitioned matrices and - between partitioned vectors shoudl be easy
• What are the mandatory modifications for the v0.1.0?
  • If there is any, which one?
  • Do I make the change in this PR?

I added, from Geoffroy review, a section mentionning the module ExpresseTreeForge as a facilitator to generate the $U_i$.

I am working on the code cells that do not produce results.

[dpo]

Let's discuss the PBFGS, etc., structures in an issue.

It would be great to implement arithmetic operations between element vectors: please open an issue.

We should register v0.1 first all of that.

[paraynaud]

Do I trigger the registrator bot?

[dpo]

First, merge approved PRs. Then, yes.