Feature Reduction Helper

GitHub Link to Code.

Feature reduction helper for data reduction operations.

This module provides utilities for processing data reduction results, handling retention warnings, and managing reduced feature metadata in the FeatureManager.

class mdxplain.feature.helper.feature_reduction_helper.FeatureReductionHelper

Helper class for feature data reduction operations.

Provides static methods for processing reduction results, handling retention rate warnings, and managing reduced feature metadata.

static process_reduction_results(feature_data: FeatureData, results: Dict[str, Any], threshold_min: int | float | None, threshold_max: int | float | None, metric: str) → None

Process and store reduction results for a single trajectory FeatureData object.

Parameters

feature_dataFeatureData: FeatureData object to update with reduction results
resultsDict[str, Any]: Results dictionary from compute_dynamic_values
threshold_minfloat or None: Minimum threshold used for context in warnings
threshold_maxfloat or None: Maximum threshold used for context in warnings
metricstr: Metric name used for context in warnings

Returns

None: Updates feature_data with reduction results and prints warnings

Examples

>>> FeatureReductionHelper.process_reduction_results(
...     feature_data, results, 0.1, 0.9, "cv"
... )
Now using reduced data. Data reduced from (1000, 500) to (1000, 45). (9.0% retained).

static reset_reduction(pipeline_data, feature_key: str) → None

Reset data reduction and return to using full original data.

Parameters

pipeline_dataPipelineData: Pipeline data object to reset
feature_keystr: Feature key to reset reduction for

Returns

None: Resets reduced data and prints summary

Examples

>>> FeatureReductionHelper.reset_reduction(pipeline_data, "distances")
Reset reduction: Now using full data (1000, 500). (Data was reduced to (1000, 45), 9.0%)

static check_reduction_state(pipeline_data, feature_key: str) → bool

Check if feature has reduced data.

Parameters

pipeline_dataPipelineData: Pipeline data object to check
feature_keystr: Feature key to check

Returns

bool: True if feature has reduced data, False otherwise

Examples

>>> has_reduction = FeatureReductionHelper.check_reduction_state(
...     pipeline_data, "distances"
... )