11.5 Tool: tf-explain tbd

tf-explain implements interpretability methods as Tensorflow 2.x callbacks to ease neural network’s understanding. See Introducing tf-explain, Interpretability for Tensorflow 2.0

Documentation: https://tf-explain.readthedocs.io

The code can be found at GitHub

tf-explain:

  • Implements interpretability methods as Tensorflow 2.x callbacks
  • Several methods to look into different aspects
    • Activations Visualization
    • Vanilla Gradients
    • Gradients*Inputs
    • Occlusion Sensitivity
    • Grad CAM (Class Activation Maps)
    • SmoothGrad
    • Integrated Gradients

Vanilla Gradients

Visualize gradients on the inputs towards the decision.

Figure from https://tf-explain.readthedocs.io/en/latest/methods.html

Gradients*Inputs

Variant of Vanilla Gradients ponderating gradients with input values.

Figure from https://tf-explain.readthedocs.io/en/latest/methods.html

Occlusion Sensitivity

Visualize how parts of the image affects neural network’s confidence by occluding parts iteratively

Figure from https://tf-explain.readthedocs.io/en/latest/methods.html

Grad CAM

Visualize how parts of the image affects neural network’s output by looking into the activation maps

Figure from https://tf-explain.readthedocs.io/en/latest/methods.html

SmoothGrad

Visualize stabilized gradients on the inputs towards the decision.

Figure from https://tf-explain.readthedocs.io/en/latest/methods.html

Integrated Gradients

Visualize an average of the gradients along the construction of the input towards the decision.

Figure from https://tf-explain.readthedocs.io/en/latest/methods.html

https://arxiv.org/abs/1611.05418