no code implementations • 24 Feb 2020 • Mathilde Guillemot, Catherine Heusele, Rodolphe Korichi, Sylvianne Schnebert, Liming Chen
The relevance spreads backward from the last to the first layer of the Deep Neural Network.