AFL.double_agent.Boundary module#

Boundary analysis tools for phase diagram and compositional spaces.

This module provides tools for analyzing and comparing boundaries in phase diagrams and compositional spaces. It includes functionality for creating concave hulls around phase regions and computing similarity scores between different boundary sets.

Key features: - Concave hull generation for phase regions - Boundary comparison metrics - Support for 2D compositional spaces - Visualization tools for boundary analysis - Handling of multi-phase systems

class AFL.double_agent.Boundary.BoundaryScore(gt_hull_variable: str, al_hull_variable: str, output_prefix: str = 'boundary_score', name: str = 'BoundaryScore', plot: bool = False)#

Bases: PipelineOp

Computes similarity scores between two sets of phase boundaries.

This class calculates metrics comparing the boundaries of phase regions between two different datasets (e.g., ground truth vs. predicted phases). It provides both numerical scores and optional visualization.

Parameters:

  • gt_hull_variable (str) – Name of variable containing ground truth hull data

  • al_hull_variable (str) – Name of variable containing comparison hull data

  • output_prefix (str, default="boundary_score") – Prefix to use for output variable names

  • name (str, default="BoundaryScore") – Name to use when added to a Pipeline

  • plot (bool, default=False) – Whether to create visualizations of the boundary comparisons

calculate(dataset: Dataset) → Self#

Calculate boundary similarity scores between hull sets.

Computes mean and standard deviation of distances between corresponding points on hull boundaries. Optionally creates visualizations of the comparison.

Parameters:

dataset (xr.Dataset) – Dataset containing both sets of hulls to compare

Returns:

The BoundaryScore instance with computed scores

Return type:

Self

class AFL.double_agent.Boundary.ConcaveHull(input_variable: str, output_prefix: str, hull_tracing_ratio: float = 0.2, drop_phases: List | None = None, component_dim: str = 'components', label_variable: str | None = None, segmentize_length: float = 0.025, allow_holes: bool = False, name: str = 'ConcaveHull')#

Bases: PipelineOp

Creates concave hulls around points in 2D space.

This class generates concave hulls (alpha shapes) around sets of points, useful for identifying phase boundaries in compositional spaces. It can handle multiple phases and provides options for hull generation parameters.

Parameters:

  • input_variable (str) – The name of the variable containing the points to create hulls around

  • output_prefix (str) – Prefix to use for output variable names

  • hull_tracing_ratio (float, default=0.2) – Controls the concavity of the hull. Lower values create tighter hulls

  • drop_phases (Optional[List], default=None) – List of phase labels to exclude from hull generation

  • component_dim (str, default="components") – Name of the dimension containing component coordinates

  • label_variable (Optional[str], default=None) – Name of variable containing phase labels. If None, all points are treated as one phase

  • segmentize_length (float, default=0.025) – Length to use when segmenting hull boundaries for smoother visualization

  • allow_holes (bool, default=False) – Whether to allow holes in the generated hulls

  • name (str, default="ConcaveHull") – Name to use when added to a Pipeline

calculate(dataset: Dataset) → Self#

Generate concave hulls for each phase in the dataset.

Creates concave hulls around points grouped by their phase labels. Handles 2D data only and includes options for hull parameters.

Parameters:

dataset (xr.Dataset) – The input dataset containing points and optional phase labels

Returns:

The ConcaveHull instance with generated hulls

Return type:

Self

Raises:

ValueError – If the input data is not 2D