Should accept Mapping as param_distributions in OptunaSearchCV

[yu9824]

Expected behavior

Should accept Mapping as param_distributions in OptunaSearchCV.

A typing of param_distributions is Mapping;

optuna-integration/optuna_integration/sklearn/sklearn.py

Line 727 in 8050e12

param_distributions: Mapping[str, distributions.BaseDistribution],

However, if we assign Mapping other than dict (e.g. MappingProxyType), it throws an error.

Environment

• Optuna version:3.5.0
• Optuna Integration version:3.5.0
• Python version:3.11.6
• OS:macOS-14.7-arm64-arm-64bit

Error messages, stack traces, or logs

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Cell In[1], line 6
      3 import optuna
      4 from sklearn.cross_decomposition import PLSRegression
----> 6 ocv = optuna.integration.OptunaSearchCV(
      7     PLSRegression(),
      8     param_distributions=MappingProxyType(
      9         dict(n_components=optuna.distributions.IntDistribution(1, 10))
     10     ),
     11 )

File ~/miniforge3/envs/py311/lib/python3.11/site-packages/optuna/_experimental.py:114, in experimental_class.<locals>.decorator.<locals>.wrapper.<locals>.wrapped_init(self, *args, **kwargs)
    103 @functools.wraps(_original_init)
    104 def wrapped_init(self: Any, *args: Any, **kwargs: Any) -> None:
    105     warnings.warn(
    106         "{} is experimental (supported from v{}). "
    107         "The interface can change in the future.".format(
   (...)
    111         stacklevel=2,
    112     )
--> 114     _original_init(self, *args, **kwargs)

File ~/miniforge3/envs/py311/lib/python3.11/site-packages/optuna/integration/sklearn.py:731, in OptunaSearchCV.__init__(self, estimator, param_distributions, cv, enable_pruning, error_score, max_iter, n_jobs, n_trials, random_state, refit, return_train_score, scoring, study, subsample, timeout, verbose, callbacks)
    728 _imports.check()
    730 if not isinstance(param_distributions, dict):
--> 731     raise TypeError("param_distributions must be a dictionary.")
    733 # Rejecting deprecated distributions as they may cause cryptic error
    734 # when cloning OptunaSearchCV instance.
    735 # https://github.com/optuna/optuna/issues/4084
    736 for key, dist in param_distributions.items():

TypeError: param_distributions must be a dictionary.

Steps to reproduce

from types import MappingProxyType

import optuna
from sklearn.cross_decomposition import PLSRegression

ocv = optuna.integration.OptunaSearchCV(
    PLSRegression(),
    param_distributions=MappingProxyType(
        dict(n_components=optuna.distributions.IntDistribution(1, 10))
    ),
)

Additional context (optional)

No response