May 11, 2020
Nevergrad is a derivative-free optimization platform gathering both a wide range of optimization methods and a wide range of test functions to evaluate them upon. Some of these functions have very particular structures which standard methods are not able to use. The most recent feature of Nevergrad is the ability to conveniently define a search domain, so that many algorithms in Nevergrad can automatically rescale variables and/or take into account their possibly logarithmic nature or their discrete nature, but also take into account any user-defined mutation or recombination operator. Since many problems are efficiently solved using specific operators, Nevergrad therefore now enables using specific operators within generic algorithms: the underlying structure of the problem is user-defined information that several families of optimization methods can use and benefit upon. We explain how this API can help analyze optimization methods and how to use it for the optimization of a structured Photonics physical testbed, and show that this can produce significant improvements.
Written by
Jérémy Rapin
Daniel Haziza
Olivier Teytaud
Antoine Moreau
Emmanuel Centeno
Pauline Bennet
Publisher
Evolutionary Computation Software Systems Workshop at GECCO
November 28, 2022
Nicolas Ballas, Bernhard Schölkopf, Chris Pal, Francesco Locatello, Li Erran, Martin Weiss, Nasim Rahaman, Yoshua Bengio
November 28, 2022
November 27, 2022
Andrea Tirinzoni, Aymen Al Marjani, Emilie Kaufmann
November 27, 2022
November 23, 2022
Tal Hassner, Cuong N. Nguyen, Cuong V. Nguyen, Lam Si Tung Ho, Vu Dinh
November 23, 2022
November 16, 2022
Kushal Tirumala, Aram H. Markosyan, Armen Aghajanyan, Luke Zettlemoyer
November 16, 2022
Foundational models
Latest news
Foundational models