Peering Inside the Black Box of Machine Learning for Atmospheric Science

Abstract:

Machine learning has recently proven very successful at a wide variety of atmospheric science problems. However, it is perceived to be a "black box" among many scientists, thus limiting its use and trust. In this talk, we overview methods for peering inside the black box for atmospheric science phenomena. We present methods for both traditional ML and deep learning and demonstrate their use on a variety of high-impact weather.