signxai.tf_signxai package

Subpackages

• signxai.tf_signxai.methods_impl package
  • Subpackages
    • signxai.tf_signxai.methods_impl.innvestigate package
      • Subpackages
      • Submodules
      • signxai.tf_signxai.methods_impl.innvestigate.layers module
      • Module contents
  • Submodules
  • signxai.tf_signxai.methods_impl.grad_cam module
    • GradCAM
      • GradCAM.__init__()
      • GradCAM.compute_heatmap()
    • calculate_grad_cam_relevancemap_timeseries()
    • calculate_grad_cam_relevancemap()
  • signxai.tf_signxai.methods_impl.guided_backprop module
    • build_guided_model()
    • guided_backprop_on_guided_model()
  • signxai.tf_signxai.methods_impl.signed module
    • calculate_sign_mu()
  • Module contents

Submodules

signxai.tf_signxai.methods module

Refactored TensorFlow explanation methods with a unified execution entry point.

signxai.tf_signxai.methods.register_method(name)

Decorator to register a method implementation.

signxai.tf_signxai.methods.execute(model, x, parsed_method: Dict[str, Any], **kwargs) → ndarray

Executes the specified XAI method for TensorFlow.

signxai.tf_signxai.methods_family module

TensorFlow integration for Method Family Architecture. This module provides an alternative entry point that uses the new family-based approach.

signxai.tf_signxai.methods_family.calculate_relevancemap_with_families(method: str, x: ndarray, model, neuron_selection: int, **kwargs)

Calculate relevance map using the Method Family Architecture. Falls back to original wrappers if needed.

Parameters:

• method – The XAI method to use

• x – Input data as numpy array

• model – TensorFlow/Keras model (without softmax)

• neuron_selection – Target class/neuron index

• **kwargs – Additional method-specific parameters

Returns:

Relevance map as numpy array

Module contents

signxai.tf_signxai.validate_input(*args, **kwargs)

Dummy validate_input function. Does nothing.

signxai.tf_signxai.calculate_relevancemap(method: str, x: ndarray, model, neuron_selection: int, **kwargs)

Calculates the relevance map for a given input and model using the specified TensorFlow-based method.

Parameters:

• method (str) – The XAI method to use (e.g., “gradient”, “lrp.epsilon”).

• x (np.ndarray) – The input data (e.g., image) as a NumPy array.

• model – The TensorFlow/Keras model (must be a model without softmax for many methods).

• neuron_selection (int) – The index of the output neuron for which to generate the explanation.

• **kwargs – Additional arguments specific to the chosen XAI method.

Returns:

The calculated relevance map.

Return type:

np.ndarray

Raises:

ValueError – If the method is not supported or if inputs are invalid.