search

mathurinm / andersoncd

This code is no longer maintained. The codebase has been moved to https://github.com/scikit-learn-contrib/skglm. This repository only serves to reproduce the results of the AISTATS 2021 paper "Anderson acceleration of coordinate descent" by Quentin Bertrand and Mathurin Massias.
BSD 3-Clause "New" or "Revised" License
18 stars 6 forks source link

ENH : add a Datafit class #42

Closed mathurinm closed 2 years ago

mathurinm commented 3 years ago

implementing grad_j(self, w, X, Xw, j), lipshitz_constant, value

For Quadratic we need to store Xj_t_y and it does not seem straightforward

mathurinm commented 3 years ago

We can call datafit.initialize(X, y) at the beginning of solver_path(), which would set whatever attributes are needed for this datafit (i.e. only Quadratic would store Xt_y). these attributes would be later accessed by datafit.grad_j, and we do not need to pass 10000 arguments to grad_j.

Sounds reasonable @QB3 ?

And we could also call penalty.initialize(X, y) for penalties such as Cel0 which depend on X

mathurinm commented 2 years ago

solved by #29