Army Creates Neural Network Metric for Testing Algorithm Reliability
The Army has developed new neural network performance metrics to measure the reliability of next-generation artificial intelligence and machine learning algorithms. The metrics will supposedly help deep neural networks apply safe techniques for command and control systems, precision fire and decision support systems.
DNNs can typically only make predictions within their scope of training, becoming unreliable when presented with wildly different information.
The Army plans to apply the technology in areas such as cybersecurity and the internet of things, MeriTalk reported Wednesday.
Brian Jalaian, a scientist at the Combat Capabilities Development Command's Army Research Laboratory, said the project may enhance machine learning algorithms based on visual imagery DNNs. “This opens a new research opportunity to create the next generation of algorithms that are robust and resilient,” Jalaian said.
Category: Popular Voices
Tags: artificial intelligence Brian Jalaian CCDC Deep Neural Networks DNNs machine learning MeriTalk Popular Voices U.S. Army
