.. _auto_doe_sampling_sequential_reduction:

*SAMPLING DOMAIN REDUCER*
=========================

Narrows the sampling domain around a known optimum by shrinking variable bounds according to a chosen reduction strategy (distance-based, direct scaling, or iso-parametric). Optionally drops low-sensitivity variables to constants using feature-importance rankings, accelerating convergence in sequential DOE workflows.

When to use
-----------

Tagged: ``doe``, ``domain reduction``, ``feature importance``, ``optimization loop``, ``sampling``, ``sequential``, ``variable bounds``.

Inputs
------

.. list-table::
   :header-rows: 1
   :widths: 20 20 20 20 20 20

   * - Label
     - ID
     - Type
     - Default
     - Required
     - Description
   * - Current Variables
     - current_variables
     - dataset
     - —
     - ✓
     - Dataset describing the current design variables including their names, types, lower/upper bounds, and any constant flags — serves as the sampling domain to be reduced.
   * - Current Optimum
     - current_optimum
     - dataset
     - —
     - ✓
     - Dataset containing the current best (optimum) variable values around which the new, tighter bounds will be centered.
   * - Percentage Reduction
     - percentage_reduction
     - scalar
     - 80
     - ✓
     - Fraction by which to shrink the variable bounds around the optimum; accepts a percentage (e.g., 80 means 80 %) or a ratio (e.g., 0.8) — defaults to 80.
   * - Variable Names
     - variable_names
     - text
     - —
     - 
     - Optional comma-separated list of variable names to restrict reduction to a specific subset; leave blank to apply reduction to all variables.
   * - Reduction Type
     - reduction_type
     - select
     - distance_from_min_max_max
     - 
     - Strategy used to compute the new bounds: 'distance_from_min_max_max' shrinks the gap between optimum and original min/max (default); 'scale_optimum_value' scales the optimum value directly; 'scale_isoparametric' uses iso-parametric position; 'scale_optimum_value_within_minmax' scales while clamping to original limits.
   * - Feature Importance
     - feature_importance
     - dataset
     - —
     - 
     - Optional dataset of per-variable sensitivity/importance scores; when provided, variables below the importance threshold are frozen to constants rather than being sampled.
   * - Feature Importance Threshold
     - feature_importance_drop_threshold
     - scalar
     - 0.8
     - 
     - Percentile/score cutoff (0–1 scale, default 0.8) below which a variable's sensitivity is considered negligible and the variable is converted to a constant.
   * - Convert Boundary To Constant Type
     - expansion
     - select
     - no
     - 
     - Controls behaviour when the optimum sits on a bound: 'no' leaves it unchanged (default), 'constant' converts the boundary variable to a constant, or 'expand' pushes the bound outward.

Outputs
-------

.. list-table::
   :header-rows: 1
   :widths: 20 20 20 20

   * - Label
     - ID
     - Type
     - Description
   * - New Variables
     - dataset
     - dataset
     - Updated variables dataset with reduced bounds centered around the current optimum, ready for use as the sampling domain in the next sequential DOE iteration.

Disciplines
-----------

- ai_ml.prognosis
- design_exploration.doe
- design_exploration.optimization
- design_exploration.sensitivity

.. raw:: html

   <hr style="margin-top:2em">
   <p style="font-size:11px;color:#888">
   Auto-generated from <code>platform</code> schema. Worker id: <code>doe_sampling_sequential_reduction</code>. Schema hash: <code>99b1a582442c</code>. Hand-curated docs in <code>workerexamples/</code> override this page when present.
   </p>