Leap Labs: AI Model Interpretability

At Leap Labs we have developed a novel approach to AI interpretability. Our methods are model agnostic, and don’t require access to the original training data. We can extract information on what a neural network has learned directly from the model itself, and we’re packaging this capability up into a library and web dashboard for anyone to use. We’d love to show you what we have so far with our initial support for image classification models.