Deep Learning and XAI Techniques for Anomaly Detection

2

Understanding Explainable AI

The breakthrough of ImageNet in 2012 accelerated the deep learning revolution, which evolved into the call for Explainable Artificial Intelligence (XAI) in 2015, initiated by the United States (U.S.) Defense Advanced Research Project Agency (DARPA), due to increased research interests in these opaque systems. Realizing the gap between rationalizing AI systems for human users and concerns about hidden risks for high-stakes situations, DARPA launched the XAI program in May 2017, intending to develop AI systems that end users can understand and trust.

In the past several decades, AI has evolved at an unprecedented rate to the extent that it almost requires no human intervention. While deep learning is increasingly ...