Abstract: The interpretation of intermediate representations in deep neural networks is critical for enhancing the transparency, trustworthiness, and applicability of artificial intelligence (AI) ...