o }oii@sddlmZddlmZddlZddlZddlZddlmZddlm Z ddlm Z ddlm Z ddl Z ddlZddlmZdd lmZdd lmZdd lmZdd lmZdd lmZddlmZddlmZddlmZddlmZddlm Z ddl!m"Z"ddl#m$Z$ddl%m&Z&ddl'm(Z(ddl'm)Z)e rddl*Z*e e*j+e*j,e*j-fZ.nedZ*e/e0Z1dZ2dZ3GdddeZ4d%d#d$Z5dS)&) annotations)SequenceN)Any)cast) TYPE_CHECKING)Union) _deprecated)logging)warn_experimental_argument) _LazyImport)_SearchSpaceTransform) optuna_warn)BaseDistribution)FloatDistribution)IntDistribution) BaseSampler)&_INDEPENDENT_SAMPLING_WARNING_TEMPLATE)LazyRandomState)IntersectionSearchSpace)StudyDirection) FrozenTrial) TrialStatecmaesg|=ic @seZdZdZ      d`ddddddddddad d!Zdbd"d#Zdcd)d*Zddd-d.Zeded0d1Z eded2d3Z dfd6d7Z dgd9d:Z dhd>d?Z didEdFZdjdKdLZdkdNdOZdldQdRZdmdUdVZdndXdYZdod^d_ZdS)p CmaEsSampleruh'A sampler using `cmaes `__ as the backend. Example: Optimize a simple quadratic function by using :class:`~optuna.samplers.CmaEsSampler`. .. code-block:: console $ pip install cmaes .. testcode:: import optuna def objective(trial): x = trial.suggest_float("x", -1, 1) y = trial.suggest_int("y", -1, 1) return x**2 + y sampler = optuna.samplers.CmaEsSampler() study = optuna.create_study(sampler=sampler) study.optimize(objective, n_trials=20) Please note that this sampler does not support CategoricalDistribution. However, :class:`~optuna.distributions.FloatDistribution` with ``step``, (:func:`~optuna.trial.Trial.suggest_float`) and :class:`~optuna.distributions.IntDistribution` (:func:`~optuna.trial.Trial.suggest_int`) are supported. If your search space contains categorical parameters, I recommend you to use :class:`~optuna.samplers.TPESampler` instead. Furthermore, there is room for performance improvements in parallel optimization settings. This sampler cannot use some trials for updating the parameters of multivariate normal distribution. For further information about CMA-ES algorithm, please refer to the following papers: - `N. Hansen, The CMA Evolution Strategy: A Tutorial. arXiv:1604.00772, 2016. `__ - `A. Auger and N. Hansen. A restart CMA evolution strategy with increasing population size. In Proceedings of the IEEE Congress on Evolutionary Computation (CEC 2005), pages 1769–1776. IEEE Press, 2005. `__ - `N. Hansen. Benchmarking a BI-Population CMA-ES on the BBOB-2009 Function Testbed. GECCO Workshop, 2009. `__ - `Raymond Ros, Nikolaus Hansen. A Simple Modification in CMA-ES Achieving Linear Time and Space Complexity. 10th International Conference on Parallel Problem Solving From Nature, Sep 2008, Dortmund, Germany. inria-00287367. `__ - `Masahiro Nomura, Shuhei Watanabe, Youhei Akimoto, Yoshihiko Ozaki, Masaki Onishi. Warm Starting CMA-ES for Hyperparameter Optimization, AAAI. 2021. `__ - `R. Hamano, S. Saito, M. Nomura, S. Shirakawa. CMA-ES with Margin: Lower-Bounding Marginal Probability for Mixed-Integer Black-Box Optimization, GECCO. 2022. `__ - `M. Nomura, Y. Akimoto, I. Ono. CMA-ES with Learning Rate Adaptation: Can CMA-ES with Default Population Size Solve Multimodal and Noisy Problems?, GECCO. 2023. `__ .. seealso:: You can also use `optuna_integration.PyCmaSampler `__ which is a sampler using cma library as the backend. Args: x0: A dictionary of an initial parameter values for CMA-ES. By default, the mean of ``low`` and ``high`` for each distribution is used. Note that ``x0`` is sampled uniformly within the search space domain for each restart if you specify ``restart_strategy`` argument. sigma0: Initial standard deviation of CMA-ES. By default, ``sigma0`` is set to ``min_range / 6``, where ``min_range`` denotes the minimum range of the distributions in the search space. seed: A random seed for CMA-ES. n_startup_trials: The independent sampling is used instead of the CMA-ES algorithm until the given number of trials finish in the same study. independent_sampler: A :class:`~optuna.samplers.BaseSampler` instance that is used for independent sampling. The parameters not contained in the relative search space are sampled by this sampler. The search space for :class:`~optuna.samplers.CmaEsSampler` is determined by :func:`~optuna.search_space.intersection_search_space()`. If :obj:`None` is specified, :class:`~optuna.samplers.RandomSampler` is used as the default. .. seealso:: :class:`optuna.samplers` module provides built-in independent samplers such as :class:`~optuna.samplers.RandomSampler` and :class:`~optuna.samplers.TPESampler`. warn_independent_sampling: If this is :obj:`True`, a warning message is emitted when the value of a parameter is sampled by using an independent sampler. Note that the parameters of the first trial in a study are always sampled via an independent sampler, so no warning messages are emitted in this case. restart_strategy: Strategy for restarting CMA-ES optimization when converges to a local minimum. If :obj:`None` is given, CMA-ES will not restart (default). If 'ipop' is given, CMA-ES will restart with increasing population size. if 'bipop' is given, CMA-ES will restart with the population size increased or decreased. Please see also ``inc_popsize`` parameter. .. warning:: Deprecated in v4.4.0. ``restart_strategy`` argument will be removed in the future. The removal of this feature is currently scheduled for v6.0.0, but this schedule is subject to change. From v4.4.0 onward, ``restart_strategy`` automatically falls back to ``None``, and ``restart_strategy`` will be supported in OptunaHub. See https://github.com/optuna/optuna/releases/tag/v4.4.0. popsize: A population size of CMA-ES. inc_popsize: Multiplier for increasing population size before each restart. This argument will be used when ``restart_strategy = 'ipop'`` or ``restart_strategy = 'bipop'`` is specified. .. warning:: Deprecated in v4.4.0. ``inc_popsize`` argument will be removed in the future. The removal of this feature is currently scheduled for v6.0.0, but this schedule is subject to change. From v4.4.0 onward, ``inc_popsize`` is no longer utilized within Optuna, and ``inc_popsize`` will be supported in OptunaHub. See https://github.com/optuna/optuna/releases/tag/v4.4.0. consider_pruned_trials: If this is :obj:`True`, the PRUNED trials are considered for sampling. .. note:: Added in v2.0.0 as an experimental feature. The interface may change in newer versions without prior notice. See https://github.com/optuna/optuna/releases/tag/v2.0.0. .. note:: It is suggested to set this flag :obj:`False` when the :class:`~optuna.pruners.MedianPruner` is used. On the other hand, it is suggested to set this flag :obj:`True` when the :class:`~optuna.pruners.HyperbandPruner` is used. Please see `the benchmark result `__ for the details. use_separable_cma: If this is :obj:`True`, the covariance matrix is constrained to be diagonal. Due to reduce the model complexity, the learning rate for the covariance matrix is increased. Consequently, this algorithm outperforms CMA-ES on separable functions. .. note:: Added in v2.6.0 as an experimental feature. The interface may change in newer versions without prior notice. See https://github.com/optuna/optuna/releases/tag/v2.6.0. with_margin: If this is :obj:`True`, CMA-ES with margin is used. This algorithm prevents samples in each discrete distribution (:class:`~optuna.distributions.FloatDistribution` with ``step`` and :class:`~optuna.distributions.IntDistribution`) from being fixed to a single point. Currently, this option cannot be used with ``use_separable_cma=True``. .. note:: Added in v3.1.0 as an experimental feature. The interface may change in newer versions without prior notice. See https://github.com/optuna/optuna/releases/tag/v3.1.0. lr_adapt: If this is :obj:`True`, CMA-ES with learning rate adaptation is used. This algorithm focuses on working well on multimodal and/or noisy problems with default settings. Currently, this option cannot be used with ``use_separable_cma=True`` or ``with_margin=True``. .. note:: Added in v3.3.0 or later, as an experimental feature. The interface may change in newer versions without prior notice. See https://github.com/optuna/optuna/releases/tag/v3.3.0. source_trials: This option is for Warm Starting CMA-ES, a method to transfer prior knowledge on similar HPO tasks through the initialization of CMA-ES. This method estimates a promising distribution from ``source_trials`` and generates the parameter of multivariate gaussian distribution. Please note that it is prohibited to use ``x0``, ``sigma0``, or ``use_separable_cma`` argument together. .. note:: Added in v2.6.0 as an experimental feature. The interface may change in newer versions without prior notice. See https://github.com/optuna/optuna/releases/tag/v2.6.0. NTF)consider_pruned_trialsrestart_strategypopsize inc_popsizeuse_separable_cma with_marginlr_adapt source_trialsx0dict[str, Any] | Nonesigma0 float | Nonen_startup_trialsintindependent_samplerBaseSampler | Nonewarn_independent_samplingboolseed int | Nonerr str | Nonerrr r!r"r#list[FrozenTrial] | NonereturnNonecCsd|dus| dkrtjjdddd}t|dt||_||_|p'tjj |d|_ ||_ ||_ t ||_t|_||_| |_| |_| |_| |_||_|jrQd|_n |jrXd |_nd |_|jrbtd |jritd |jdurrtd |jrytd|jrtd|dur|dus|durtd|dur| rtd| r| s| rtd|jr|jrtddSdS)Nrz`restart_strategy`z4.4.0z6.0.0)named_verr_verz~ From v4.4.0 onward, `restart_strategy` automatically falls back to `None`. `restart_strategy` will be supported in OptunaHub.)r.zsepcma:zcmawm:zcma:rr r#r!r"zQIt is prohibited to pass `source_trials` argument when x0 or sigma0 is specified.zNIt is prohibited to pass `source_trials` argument when using separable CMA-ES.z]It is prohibited to pass `use_separable_cma` or `with_margin` argument when using `lr_adapt`.zMCurrently, we do not support `use_separable_cma=True` and `with_margin=True`.)r_DEPRECATION_WARNING_TEMPLATEformatr FutureWarning_x0_sigma0optunasamplers RandomSampler_independent_sampler_n_startup_trials_warn_independent_samplingr_cma_rngr _search_space_consider_pruned_trials_popsize_use_separable_cma _with_margin _lr_adapt_source_trials _attr_prefixr ValueError)selfr$r&r(r*r,r.rrrrr r!r"r#msgrNJ/home/ubuntu/.local/lib/python3.10/site-packages/optuna/samplers/_cmaes.py__init__sj     zCmaEsSampler.__init__cCs|jdSN)r? reseed_rngrLrNrNrOrROszCmaEsSampler.reseed_rngstudy'optuna.Study'trial'optuna.trial.FrozenTrial'dict[str, BaseDistribution]cCsDi}|j|D]\}}|rq t|ttfsq |||<q |SrQ)rC calculateitemssingle isinstancerr)rLrTrV search_spacer4 distributionrNrNrOinfer_relative_search_spaceSs z(CmaEsSampler.infer_relative_search_spacer]dict[str, Any]cCs||t|dkr iS||}t||jkriSt||j dd}||}|dur4|||j}|j t|j krR|j rP|j j j}td|dd|_ iS|||j}t||jkrg} |d|jD]4} | jdustJdt|tjrt| jd} n|| j} |jtjkr| jn| j } | | | fqi| | t!"|#} |$| }|D] }|j%&|j'|||q|j(j)*d d |j+}|j,-|t|tjr|.\}}|j%&|j'd|/n|.}|j0}|j%&|j'||j|1|}|S) NrT)transform_step transform_0_1z7`CmaEsSampler` does not support dynamic search space. `z$` is used instead of `CmaEsSampler`.Fz"completed trials must have a value x_for_tellri)2_raise_error_if_multi_objectivelen _get_trialsr@r rG_restore_optimizer_init_optimizer directiondimboundsrAr? __class____name___loggerwarning_get_solution_trials generationpopulation_sizevaluer\rCMAwMnparray system_attrs transformparamsrMINIMIZEappendtellpickledumpshex_split_optimizer_str_storageset_trial_system_attr _trial_idrBrngrandintnumber_rngr.asktolist_attr_key_generation untransform)rLrTrVr]completed_trialstrans optimizerind_sampler_namesolution_trials solutionstxy optimizer_stroptimizer_attrskeyr.ryrcgeneration_attr_keyexternal_valuesrNrNrOsample_relativeesf                zCmaEsSampler.sample_relativestrcC |jdS)NrqrJrSrNrNrOr z!CmaEsSampler._attr_key_generationcCr)NrrrSrNrNrO_attr_key_optimizerrz CmaEsSampler._attr_key_optimizerrdict[str, str]cs"dfddttDS)Nc3s$|] }jd|VqdS):N)r).0irrLrNrO s z7CmaEsSampler._concat_optimizer_attrs..)joinrangere)rLrrNrrO_concat_optimizer_attrss z$CmaEsSampler._concat_optimizer_attrsrcCs^t|}i}tt|tD]}|t}t|dt|}|||||jd|<q|S)Nrr)rermathceil_SYSTEM_ATTR_MAX_LENGTHminr)rLr optimizer_lenattrsrstartendrNrNrOrsz!CmaEsSampler._split_optimizer_strr 'list[optuna.trial.FrozenTrial]''CmaClass' | NonecsTt|D]#}fdd|jD}t|dkrq|}tt|SdS)Ncs"i|] \}}|jr||qSrN) startswithr)rrrsrSrNrO s  z3CmaEsSampler._restore_optimizer..r) reversedrwrZrerr}loadsbytesfromhex)rLrrVrrrNrSrOrgs    zCmaEsSampler._restore_optimizerrr rir 'CmaClass'c s.jdddf}jdddf}tj}|jdurE|jdur*|||d}n|j}|jdur?t||d}n|j}d}n4tj g|j rR tj |t jkrYdndfdd|jD} t| dkrqtdt| \}}}t|t}|jrtjdkrtd tntj||j|jjdd d ||jd S|jrtjtjtd } t j!D]0\} } t"| t#t$fsJ| j%dus| j&rd| | <q| j'| j(krd| | <q| j%| j(| j'| | <qtj)||j| ||jjdd d ||jdStj*|||j|jjdd d ||j|j+dS)Nrrrcs>g|]}|jvrt|jr|jtt|jfqSrN)state_is_compatible_search_space distributionsrxryrfloatrsrrexpected_statessignrrNrO s  z0CmaEsSampler._init_optimizer..zNo compatible source_trialszSeparable CMA-ES does not operate meaningfully on single-dimensional search spaces. The setting `use_separable_cma=True` will be ignored.i )meansigmarkr.n_max_resamplingrr)dtypegg?)rrrkstepscovr.rrr)rrrrkr.rrrr"),rkrerIr:rxr;rurrCOMPLETErDr{PRUNEDrrzrKrget_warm_start_mgdmax_EPSrFr UserWarningSepCMArBrrrErGemptyrCr enumeratevaluesr\rrsteploglowhighrtCMArH) rLrri lower_bounds upper_bounds n_dimensionrr&rsource_solutionsrrdistrNrrOrhs             zCmaEsSampler._init_optimizer param_nameparam_distributionrrcCsF|||jr||}t||jkr||||j||||SrQ)rdrArfrer@_log_independent_samplingr?sample_independent)rLrTrVrrcomplete_trialsrNrNrOr9s   zCmaEsSampler.sample_independentrc Cs*ttj||j|jjj|jjdddS)NzVdynamic search space and `CategoricalDistribution` are not supported by `CmaEsSampler`)r trial_numberindependent_sampler_name sampler_namefallback_reason)rnrorr8rr?rlrm)rLrVrrNrNrOrKsz&CmaEsSampler._log_independent_samplinglist[FrozenTrial]cCsg}|jdddD]9}|jtjkr||q |jtjkrBt|jdkrB|jrBt |j \}}|dur5q t |}||_ ||q |S)NFT)deepcopy use_cacher)rfrrrr{rreintermediate_valuesrDrrZcopyrrs)rLrTrr_rscopied_trNrNrOrfYs      zCmaEsSampler._get_trialstrialsrqcs|jfdd|DS)Ncs"g|] }|jdkr|qS)r)rwgetrrqrrNrOrps"z5CmaEsSampler._get_solution_trials..)r)rLrrqrNrrOrplsz!CmaEsSampler._get_solution_trials optuna.StudycCs|j||dSrQ)r? before_trial)rLrTrVrNrNrOrrszCmaEsSampler.before_trialrrrSequence[float] | NonecCs|j||||dSrQ)r? after_trial)rLrTrVrrrNrNrOruszCmaEsSampler.after_trial)NNrNTN)r$r%r&r'r(r)r*r+r,r-r.r/rr-rr0rr/rr)r r-r!r-r"r-r#r1r2r3)r2r3)rTrUrVrWr2rX)rTrUrVrWr]rXr2r`)r2r)rrr2r)rrr2r)rrr2r)rr rirr2r) rTrUrVrWrrrrr2r)rVrrrr2r3)rTrUr2r)rrrqr)r2r)rTrrVrr2r3) rTrUrVrWrrrrr2r3)rm __module__ __qualname____doc__rPrRr_rpropertyrrrrrgrhrrrfrprrrNrNrNrOr/sHK  W  M      _    rrr r]rXr2r-cCs>tt|j|}|t|jkot|kSSrQ)resetrCkeys intersection)rr]intersection_sizerNrNrOrs"r)rr r]rXr2r-)6 __future__rcollections.abcrrrr}typingrrrrnumpyrur<rr optuna._experimentalr optuna._importsr optuna._transformr optuna._warningsr optuna.distributionsrrroptuna.samplersroptuna.samplers._baser"optuna.samplers._lazy_random_stateroptuna.search_spaceroptuna.study._study_directionr optuna.trialrrrrrrtCmaClass get_loggerrmrnrrrrrNrNrNrOsP                      T