AFL.automation.instrument.VirtualSAS

Functions

mpl_plot_to_bytes([fig, format])

Classes

Driver(name[, defaults, overrides, ...])
LabelEncoder()	Encode target labels with value between 0 and n_classes-1.
RandomForestClassifier([n_estimators, ...])	A random forest classifier.
VirtualSAS([overrides])

class AFL.automation.instrument.VirtualSAS.VirtualSAS(overrides=None)

defaults = {'boundary_datasets': {}, 'components': [], 'fast_locate': True, 'noise': 0.0, 'reference_data': [], 'rfc_max_depth': None, 'rfc_min_samples_leaf': 1, 'rfc_min_samples_split': 2, 'rfc_n_estimators': 100, 'rfc_random_state': 42, 'sasview_models': {}, 'ternary': False}

__init__(overrides=None)

Generates smoothly interpolated scattering data via a noiseless GPR from an experiments netcdf file. Uses RandomForestClassifier for phase boundary classification.

status()

load_reference_data()

Load reference data from config into sasmodels Data1D objects.

Config format: reference_data = [

{‘q’: […], ‘I’: […], ‘dI’: […], ‘dq’: […]}, {‘q’: […], ‘I’: […], ‘dI’: […], ‘dq’: […]},

]

load_sasview_models()

Load SASView models from config.

Config format: sasview_models = {

‘phase_A’: {‘model_name’: ‘sphere’, ‘model_kw’: {‘radius’: 50, …}}, ‘phase_B’: {‘model_name’: ‘cylinder’, ‘model_kw’: {‘radius’: 20, …}},

}

validate_boundary_datasets_config()

Validate boundary_datasets config structure and dimensionality.

Returns:

True if valid, raises ValueError otherwise

Return type:

bool

Raises:

ValueError – If config structure is invalid

migrate_boundary_dataset_to_config()

Convert old boundary_dataset xarray object to new config format.

This is a migration helper for transitioning from the old shapely-based system to the new RFC-based system.

Returns:

Boundary datasets in new config format

Return type:

dict

Raises:

ValueError – If boundary_dataset is None or invalid

train_classifier(reset=True, drop_phases=None)

Train RandomForestClassifier on boundary datasets from config.

Parameters:

• reset (bool) – If True, reinitialize classifier before training

• drop_phases (list or None) – Phase labels to exclude from training

Raises:

ValueError – If boundary_datasets not configured or invalid format

locate(composition)

Predict phase membership using trained RandomForestClassifier.

Parameters:

composition (array-like) – Composition vector. Dimensionality depends on ternary mode: - ternary=True: Must be 3D (will be transformed to 2D for RFC) - ternary=False: Any dimension matching training data

Returns:

ds – Dataset containing: - ‘phase’: predicted phase label - ‘probability’: confidence score - attrs: all phase probabilities

Return type:

xr.Dataset

Notes

Use locate_with_uncertainty() to get prediction probabilities as tuple

locate_with_uncertainty(composition)

Predict phase with uncertainty estimation via predict_proba.

Parameters:

composition (array-like) – Composition vector. Dimensionality depends on ternary mode: - ternary=True: Must be 3D (will be transformed to 2D for RFC) - ternary=False: Any dimension matching training data

Returns:

• phase (str) – Predicted phase label

• probability (float) – Confidence score for predicted phase (0-1)

• all_probabilities (dict) – Probability scores for all phases {phase_label: probability}

generate(label)

Generate scattering data for a given phase label.

Parameters:

label (str) – Phase label (must exist in sasview_models config)

Returns:

ds – Dataset containing: - ‘q’: scattering vector - ‘I’: scattered intensity (with noise) - ‘I_noiseless’: scattered intensity (without noise) - ‘dI’: uncertainty - attrs: phase label, model name

Return type:

xr.Dataset

expose(*args, **kwargs)

Mimic the expose command from other instrument servers.

Returns:

ds – Combined dataset from locate() and generate() with composition info

Return type:

xr.Dataset

plot_decision_boundaries(grid_resolution=200, **kwargs)

Plot RFC decision boundaries with training data overlay.

Parameters:

grid_resolution (int) – Number of grid points per axis for decision boundary mesh

plot_hulls(**kwargs)

DEPRECATED: Use plot_decision_boundaries() instead. This wrapper is provided for backward compatibility.

plot_boundary_data(**kwargs)

Plot boundary training data in ternary projection (if ternary mode enabled)