• AI2.JPG

 

Nowadays neural networks are often used to analyse complex data - for example, in order to discover information about diseases in genetic information. Ultimately, though, no one knows how these networks exactly decide. Researchers at Fraunhofer HHI and TU Berlin have developed a software that enables them to look into these black boxes. It allows researchers to watch neural networks “think” and to analyse how they work. The demo highlights pixels that were most influential on decision-making.

The Visual-Question-Answering-System (VQA) is able to analyse the content of pictures and answer content-related questions. Again pixels are highlighted, which the algorithm used to recognize the content. This enables researchers to better understand the functioning of the AI.  

http://s.fhg.de/machinelearning  

 

Diese Website speichert Informationen über Ihren Besuch in sogenannten Cookies. Durch die Nutzung dieser Webseite erklären Sie sich mit der Verwendung von Cookies einverstanden. Mehr zum Datenschutz finden Sie in unserer Datenschutzerklärung. Ich akzeptiere Cookies von www.ciniq.de