Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Optimization methods specifically for RL #1273

Closed
wants to merge 12 commits into from
Closed

Optimization methods specifically for RL #1273

Show file tree
Hide file tree
Changes from 11 commits
Commits
Show all changes
12 commits
Select commit Hold shift + click to select a range
5113452
Update optimizerlib.py
teytaud Oct 22, 2021
1199cea
record
teytaud Oct 22, 2021
0a0b9e4
rl
teytaud Oct 22, 2021
cb4ac62
Update recorded_recommendations.csv
teytaud Oct 22, 2021
7a23928
Update optimizerlib.py
teytaud Oct 22, 2021
3a35787
better recom for chaining: follow the last optimization method
teytaud Oct 22, 2021
9cee7cf
fix_typing
teytaud Oct 22, 2021
296f64f
Update optimizerlib.py
teytaud Oct 23, 2021
f069c6d
Update recorded_recommendations.csv
teytaud Oct 23, 2021
62eb584
fix
teytaud Oct 23, 2021
eb76921
Merge branch 'main' into for_rl
teytaud Nov 4, 2021
090ce00
Merge branch 'main' of github.com:facebookresearch/nevergrad into for_rl
teytaud Nov 8, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 10 additions & 0 deletions nevergrad/optimization/optimizerlib.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2059,6 +2059,9 @@ def _internal_tell_candidate(self, candidate: p.Parameter, loss: tp.FloatLoss) -
if self.num_tell < sum_budget:
opt.tell(candidate, loss)

def _internal_provide_recommendation(self) -> tp.ArrayLike:
return self.optimizers[-1]._internal_provide_recommendation() # type: ignore


class Chaining(base.ConfiguredOptimizer):
"""
Expand Down Expand Up @@ -2087,6 +2090,10 @@ def __init__(
GeneticDE = Chaining([RotatedTwoPointsDE, TwoPointsDE], [200]).set_name(
"GeneticDE", register=True
) # Also known as CGDE
MixDeterministicRL = ConfPortfolio(optimizers=[DiagonalCMA, PSO, GeneticDE]).set_name(
"MixDeterministicRL", register=True
)
SpecialRL = Chaining([MixDeterministicRL, TBPSA], ["half"]).set_name("SpecialRL", register=True)
Comment on lines +2093 to +2096
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Do we need both?

Suggested change
MixDeterministicRL = ConfPortfolio(optimizers=[DiagonalCMA, PSO, GeneticDE]).set_name(
"MixDeterministicRL", register=True
)
SpecialRL = Chaining([MixDeterministicRL, TBPSA], ["half"]).set_name("SpecialRL", register=True)
SpecialRL = Chaining([ConfPortfolio(optimizers=[DiagonalCMA, PSO, GeneticDE]), TBPSA], ["half"]).set_name("SpecialRL", register=True)

ChainCMAPowell = Chaining([CMA, Powell], ["half"]).set_name("ChainCMAPowell", register=True)
ChainCMAPowell.no_parallelization = True # TODO make this automatic
ChainMetaModelSQP = Chaining([MetaModel, SQP], ["half"]).set_name("ChainMetaModelSQP", register=True)
Expand Down Expand Up @@ -2760,3 +2767,6 @@ def __init__(
base_optimizer: base.OptCls = NGOpt,
) -> None:
super().__init__(_MSR, locals())


NGOptRL = SpecialRL
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'd rather we keep one name only if possible, why have two names for now?
As usual, adding more names means more breaking changes down the roads, so I prefer being careful.

3 changes: 3 additions & 0 deletions nevergrad/optimization/recorded_recommendations.csv
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -120,6 +120,7 @@ MilliCMA,0.0010125155,-0.0009138806,-0.0010295559,0.0012098418,,,,,,,,,,,,
MiniDE,0.8273276988,-1.2921051963,-0.4797521288,0.2138608624,0.7088815721,0.7346249014,-2.6392592028,-1.0729615222,,,,,,,,
MiniLhsDE,-0.0313128807,0.2738703026,-0.1988242191,0.9942001938,0.7167500893,-0.0350394443,-1.5341684983,-0.3039246928,,,,,,,,
MiniQrDE,-0.2025746195,-0.8778768047,-1.2504657435,0.6265108481,0.4934247309,0.6448108695,-0.3573249779,-1.6986947217,,,,,,,,
MixDeterministicRL,-0.3375952501,-0.585268278,-0.1149199408,2.2418177944,,,,,,,,,,,,
MixES,1.1400386808,0.3380024444,0.4755144618,2.6390460807,0.6911075733,1.111235567,-0.2576843178,-1.1959512855,,,,,,,,
MultiCMA,1.4855013085,-1.5953064496,-0.7871164493,-0.4908938162,,,,,,,,,,,,
MultiDiscrete,0.0,0.0,0.0,1.095956118,,,,,,,,,,,,
Expand Down Expand Up @@ -199,6 +200,8 @@ ScrHaltonSearchPlusMiddlePoint,-1.1503493804,1.2206403488,-0.8416212336,1.067570
ScrHammersleySearch,1.3829941271,-0.318639364,-1.2206403488,1.7506860713,,,,,,,,,,,,
ScrHammersleySearchPlusMiddlePoint,-1.2815515655,0.0,0.4307272993,0.8416212336,,,,,,,,,,,,
Shiwa,0.0,-0.3451057176,-0.1327329683,1.9291307781,,,,,,,,,,,,
SpecialRL,-0.3375952501,-0.585268278,-0.1149199408,2.2418177944,,,,,,,,,,,,
NGOptRL,-0.3375952501,-0.585268278,-0.1149199408,2.2418177944,,,,,,,,,,,,
SparseDiscreteOnePlusOne,0.7531428339,0.0,0.0,0.0,,,,,,,,,,,,
SparseDoubleFastGADiscreteOnePlusOne,0.0,0.0,0.0,0.0942747145,,,,,,,,,,,,
StupidRandom,-1.1543602352,-2.2133334794,-1.6817565104,-1.7880942511,,,,,,,,,,,,
Expand Down