Disentangling Influence: Using disentangled representations to audit model predictions

Feature influence can also be expressed in aggregate over the training or test data or locally with respect to a single point.