Along with the adaptation of machine learning techniques for domain sciences, computing advances and new technologies such as edge computing platforms have allowed software-defined sensors, which are reprogrammable software programs for processing rich data to be moved to the edge to process the sensor data in situ and provide more comprehensive sensing of the surrounding environment where they are deployed. Additionally, using edge computing platforms has the advantage that sensor configurations can be modified automatically in real-time derived from local artificial intelligence. In this talk, we will present some examples of utilizing supervised machine learning techniques on edge computing platforms for providing high-dimensional information that helps domain scientists understand the environment better. Additionally, we will explore the need for explainable artificial intelligence and the edge-to-cloud / HPC continuum of computing paradigm from training and optimizing machine learning models to supporting wide-area analysis based on local information aggregation. We will also examine how local AI can assist in automatic sensor reconfiguration to steer the measurements and improve the acquired data.

ADD TO CALENDAR / View Recording