A new technique called 'concept whitening' promises to provide neural network interpretability

Open in new window