otaf.optimization package

Module contents

Tool to track optimization history and manage constraints.

class otaf.optimization.OptimizationTracker(bounds=None, constraint_tolerance=0.0001, precision_decimals=8)[source]

Bases: object

High-performance storage for optimization loops.

Uses byte-hashing for O(1) lookups during execution, and exports to Pandas for post-processing and filtering.

Parameters:

  • bounds (Bounds, optional) – The optimization bounds object exposing a .residual(x) method. Default is None.

  • constraint_tolerance (float, optional) – The maximum allowed tolerance for constraint violations. Default is 1e-4.

  • precision_decimals (int, optional) – The number of decimals to round coordinate values before byte-hashing. Default is 8.

history

Nested dictionary tracking evaluation points by experiment key and point hash.

Type:

dict

precision_decimals

Decimal precision for coordinate rounding.

Type:

int

bounds

Optimization bounds constraint definition.

Type:

Bounds or None

constraint_tolerance

Tolerance threshold for constraint validation.

Type:

float

filter_points(exp_key=None, source=None, bounds_respected=None, constraints_respected=None)[source]

Filter tracking records against targeted properties via DataFrame.

Parameters:

  • exp_key (Any, optional) – Filters historical points to a single tracked experiment key sequence. Default is None.

  • source (str, optional) – Filters matching points by execution string label. Default is None.

  • bounds_respected (bool, optional) – Filters items based on validation boundary constraints compliance. Default is None.

  • constraints_respected (bool, optional) – Filters items based on tolerance limit boundary validation. Default is None.

Returns:

A targeted slice of history matching all provided validation conditions.

Return type:

pd.DataFrame

get_data(exp_key, x)[source]

Retrieve internal stored attributes for a point if existing.

Parameters:

  • exp_key (Any) – The identification key for the tracking sequence.

  • x (array_like) – The exact coordinate values to fetch.

Returns:

The dictionary of matching parameters, or None if the point cannot be found.

Return type:

dict or None

to_dataframe(exp_key=None)[source]

Convert the internal dictionary to a Pandas DataFrame.

Replaces the old get_all_data and avoids loop bottlenecks.

Parameters:

exp_key (Any, optional) – The specific experiment key to extract. If None, flattens and extracts all stored items across tracking sequences. Default is None.

Returns:

Flattened historical evaluation frame containing structural parameter rows.

Return type:

pd.DataFrame

update_constraint_data(exp_key, x, constraints)[source]

Update constraint evaluation metrics and assess tolerance breaches.

Parameters:

  • exp_key (Any) – The identifier of the active experiment.

  • x (array_like) – The coordinates of the evaluated point.

  • constraints (array_like) – Array or values representing current constraint evaluation outputs.

Return type:

None

update_objective_data(exp_key, x, fp_gld, fp_slack, gld_params, failure_slack, source='local')[source]

Update objective function metrics for a specific point.

Parameters:

  • exp_key (Any) – The identifier of the active experiment.

  • x (array_like) – The coordinates of the evaluated point.

  • fp_gld (float) – Failure probability obtained from the GLD distribution approximation (converted to NaN if input is NaN).

  • fp_slack (float) – Failure probability obtained from number of failed points in sample.

  • gld_params (array_like) – Parameters of the GLD distribution.

  • failure_slack (float) – The slack value used to define failure.

  • source (str, optional) – The metadata tracking origin string of the evaluation. Default is “local”.

Return type:

None