add static doc content for scipy optimizers

Author: SimonBlanke

docs/source/api_reference/optimizers.rst

@@ -115,6 +115,57 @@ These optimizers provide an interface to Optuna's optimization algorithms.
     NSGAIIIOptimizer
     QMCOptimizer
 
+Scipy-Based Optimizers
+-----------------------
+
+These optimizers provide an interface to scipy.optimize algorithms.
+They support **continuous parameter spaces only** (tuples). For discrete
+or categorical parameters, use Optuna or GFO backends.
+
+Global Optimizers
+~~~~~~~~~~~~~~~~~
+
+Global optimization algorithms for finding global optima.
+
+.. autosummary::
+    :toctree: auto_generated/
+    :template: class.rst
+
+    ScipyDifferentialEvolution
+    ScipyDualAnnealing
+    ScipyBasinhopping
+    ScipySHGO
+    ScipyDirect
+
+Local Optimizers
+~~~~~~~~~~~~~~~~
+
+Derivative-free local optimization algorithms.
+
+.. autosummary::
+    :toctree: auto_generated/
+    :template: class.rst
+
+    ScipyNelderMead
+    ScipyPowell
+
+SMAC-Based Optimizers
+---------------------
+
+These optimizers provide an interface to SMAC3's Bayesian optimization algorithms.
+SMAC3 (Sequential Model-based Algorithm Configuration) uses Random Forest or
+Gaussian Process surrogate models for efficient hyperparameter optimization.
+
+Install with: ``pip install smac`` or ``pip install hyperactive[smac]``
+
+.. autosummary::
+    :toctree: auto_generated/
+    :template: class.rst
+
+    SmacRandomForest
+    SmacGaussianProcess
+    SmacRandomSearch
+
 Scikit-Learn Style
 -------------------
 

docs/source/user_guide/optimizers.rst

@@ -35,6 +35,9 @@ The best optimizer depends on your problem characteristics:
    * - Small search space
      - ``GridSearch``
      - Exhaustive coverage when feasible
+   * - Continuous params only
+     - ``ScipyDifferentialEvolution``, ``ScipyDualAnnealing``
+     - Well-tested scipy implementations
 
 
 Optimizer Categories
@@ -264,6 +267,39 @@ Example with Optuna TPE:
    :end-before: # [end:optuna_tpe]
 
 
+Scipy Backend
+^^^^^^^^^^^^^
+
+Scipy optimizers are well-tested implementations for **continuous parameter spaces only**.
+They do not support discrete or categorical parameters.
+
+Available scipy optimizers:
+
+- **Global**: ``ScipyDifferentialEvolution``, ``ScipyDualAnnealing``,
+  ``ScipyBasinhopping``, ``ScipySHGO``, ``ScipyDirect``
+- **Local**: ``ScipyNelderMead``, ``ScipyPowell``
+
+Example with Scipy Differential Evolution:
+
+.. code-block:: python
+
+   from hyperactive.opt.scipy import ScipyDifferentialEvolution
+
+   optimizer = ScipyDifferentialEvolution(
+       param_space={
+           "learning_rate": (0.0001, 0.1),
+           "momentum": (0.5, 0.99),
+       },
+       n_iter=100,
+       strategy="best1bin",
+       random_state=42,
+       experiment=experiment,
+   )
+   best_params = optimizer.solve()
+
+For discrete or categorical parameters, use GFO or Optuna backends instead.
+
+
 Optimizer Configuration
 -----------------------
 

docs/source/user_guide/search_spaces.rst

@@ -82,6 +82,34 @@ For parameters that vary continuously, use NumPy to create arrays:
     though Hyperactive accepts both formats.
 
 
+Scipy Backend: Tuple Ranges
+^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+Scipy optimizers use a different format: tuples ``(low, high)`` for truly continuous
+optimization (not discretized):
+
+.. code-block:: python
+
+    from hyperactive.opt.scipy import ScipyDifferentialEvolution
+
+    # Scipy uses tuples for continuous bounds
+    param_space = {
+        "learning_rate": (0.0001, 0.1),   # Continuous range
+        "momentum": (0.5, 0.99),           # Continuous range
+    }
+
+    optimizer = ScipyDifferentialEvolution(
+        param_space=param_space,
+        n_iter=100,
+        experiment=experiment,
+    )
+
+.. warning::
+
+    Scipy optimizers do **not** support lists or categorical values.
+    Use GFO or Optuna backends for discrete/categorical parameters.
+
+
 Scale-Appropriate Spacing
 -------------------------