Charlie Vanaret
banner
cvanaret.bsky.social
Charlie Vanaret
@cvanaret.bsky.social
Mathematical optimizer at Argonne/Zuse Institute Berlin
My nonlinear optimization solver Uno unifies SQP and barrier methods
Bassist/landscape photographer

🌐 github.com/cvanaret/Uno
If you live in Germany, get high-quality prints of the ARKTIS pictures here: www.etsy.com/de/shop/Char...
www.etsy.com
May 2, 2025 at 2:38 PM
@schmaidt.bsky.social in case you're around... ;)
April 16, 2025 at 2:02 PM
Ksenia Bestuzheva (SCIP)

www.zib.de/members/best...
Bestuzheva, Ksenia, Dr. | zib.de
www.zib.de
March 19, 2025 at 7:51 AM
What about Ivet Galabova (HiGHS)?
March 18, 2025 at 6:46 PM
Definitely! Here there is no cancellation. I meant that the complex step, compared to other finite difference methods, doesn't produce cancellation (and is second-order).
December 4, 2024 at 12:25 PM
That shit is amazing :) However, a single evaluation at x + i*h is required, you shouldn't subtract f(x) there. It avoids catastrophic cancellation.
December 4, 2024 at 12:16 PM