chmncc.explanations package

Submodules

chmncc.explanations.gradients module

Module which deals with the explaination we can retrieve out of the images

chmncc.explanations.gradients.compute_integrated_gradient(batch_x: Tensor, batch_blank: Tensor, model: Module) → Tensor

Integrated gradients computation Implementation taken from https://github.com/CVxTz/IntegratedGradientsPytorch/blob/main/code/mlp_gradient.py

Parameters:

• batch (batch_x [torch.Tensor] data instances) –

• zeros (batch_blank [torch.Tensor] batch of) –

• network (model [torch.nn.Module]) –

Returns:

integrated_gradients [torch.Tensor] integrated gradients

chmncc.explanations.gradients.output_gradients(inputs: Tensor, preds: Tensor) → Tensor

Compute the gradients with respect to the input

Parameters:

• tensor (preds [torch.Tensor] output) –

• tensor –

Returns:

grad [torch.Tensor] gradients of the output with respect to the input

Module contents

Explainations For now, they are simply grandients based