A research team from ORNL, Pacific Northwest National Laboratory, and Arizona State University has developed a novel method to detect out-of-distribution (OOD) samples in continual learning without forgetting the learned knowledge of preceding tasks.
