The article focuses on a study conducted by researcher Junfeng Yang and colleagues to create Deep-Xplore, a program designed to reverse learning processes of artificial intelligence system to debug them. The study tested 15 state-of-the-art neural networks and found that DeepXplore boosted the AI systems' overall accuracy by one to three percent on average after discovering thousands of bugs.
© 2001-2025 Fundación Dialnet · Todos los derechos reservados