Explainable Neural Networks

Gradient Based Approaches

Visualizing Gradients (2014)

In the paper Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps Karen Simonyan et. al. proposed two visualisation techniques, based on computing the gradient of the class score with respect to the input image.

  1. The first technique generates an image, which maximises the class score, visualising the notion of the class, captured by a ConvNet.
  2. The second technique computes a class saliency map, specific to a given image and class.