Fabian Pedregosa

Results 200 comments of Fabian Pedregosa

Similar problem here, with images displaying at different scale that supposed. In my case these are embedded as PDF. I'm using atom 1.10 and latex 0.36 ![image](https://cloud.githubusercontent.com/assets/277639/18616615/5d9e76b8-7dbf-11e6-926b-2c584f8bdf43.png)

Hitting `Control -` once solves the problem: ![image](https://cloud.githubusercontent.com/assets/277639/18616653/222f3a62-7dc0-11e6-9ae6-8a2a00770f1b.png) but hitting it more than once makes it worse: ![image](https://cloud.githubusercontent.com/assets/277639/18616660/2e4f7ed8-7dc0-11e6-8236-e5f2d7f0c40a.png) It's also strange that hitting `Control -` makes the images bigger instead...

Thanks for the response. Otherwise I think it could detect it automatically. If the first line has text but the second one is a blank line I think it would...

Cool ! I'll be offline until Monday, then I'll definitely take a look into it.

since it is specific to least squares I would call it _frank_wolfe_ls (and also because I would love a FW for arbitrary loss functions :-). Just FYI yesterday I saw...

What about using [scipy's line_search](http://docs.scipy.org/doc/scipy-0.16.0/reference/generated/scipy.optimize.line_search.html) for arbitrary loss functions and the exact one for the quadratic one?

In theory FISTA should work with constrained problems using the projection as proximal operator. Have you tried with simple ISTA=projected gradient descent to debug (trivial to implement)?