Releases: facebookresearch/nevergrad
More plots, better texts, incorporating new chaining methods (Carola*) and new wizards (NgIoh and Wiz)
More plots,
better texts (in the automatic latex creation),
incorporating new chaining methods (Carola*)
new wizards (NgIoh and Wiz)
0.11.0
Fix the Boston issue
Contributors
Assets 2
More quasi-opposite, switch to Python3.8
What's Changed
- Dagstuhloid benchmark ++ by @teytaud in #1534
- New Dagstuhloid benchmark, including quasi-opposite stuff by @teytaud in #1540
Full Changelog: 0.8.0...v0.9.0
Contributors
Assets 2
Adding the Dagstuhloid benchmark and many more metamodels
Inspired by discussions at the Dagstuhl seminar, and others.
What's Changed
- Update bench.txt by @teytaud in #1522
- Voronoi differential evolution, first draft by @teytaud in #1523
- adding another CMA and a Bipop by @teytaud in #1524
- new metamodels by @teytaud in #1517
- topology optimization artificial benchmark by @teytaud in #1525
- bfgs by @teytaud in #1532
- Update utils.py for fault tolerance by @teytaud in #1496
- Implementation of a Dagstuhloid benchmark, i.e. a benchmark inspired by discussions at Dagstuhl by @teytaud in #1531
Full Changelog: 0.7.0...v0.8.0
Contributors
Assets 2
Adding weighted MOO and various metamodels
adding weighted DE multiobjective
adding various metamodels
Cleaning, better naming.
Adding NGOptRW, presumably better than NGOpt for real-world problems.
Adding YAPBBOB, with a parameter regulating YABBOB-like problems so that the distribution of the optimum is less rotationally invariant.
Making some dependencies optional because running was becoming too complicated.
There should be no breaking change.
Adding constrained counterparts of YABBOB: yapenbbob (a few constraints), yaonepenbbob (single constraint), yamegapenbbob (many constraints).
Improvements in the photonics benchmarks.
Externalizing CompilerGym.
Making some tests less flaky.
Adding Simulated annealing and Tabu search.
Adding the NLOPT library.
Making the code more robust to Gym environments.
Adding smoothness operators for discrete optimization.
Adding NGOptRW and add constraint penalization
v0.5.0 Update test_doc.py (#1474)
0.5.0
0.4.3
This version provides a few fixes and the new multi-objective API of optimizers (you can now provide a list/array of float to tell directly). This allows fore more efficient multi-objective optimization for some optimizers (DE, NGOpt). Future work will continue to improve multi-objective capacities and aim at improving constraints management.
See CHANGELOG for details.