Quantifying Uncertainty in Machine Learning-Assisted Measurements

Like any science and engineering field, Instrumentation and Measurement (I&M) is currently experiencing the impact of the recent rise of Artificial Intelligence and in particular Machine Learning (ML). In fact, there is an intertwined relationship between the two: I&M is used to collect data, which are then used to train ML models, which in turn are used in I&M systems. The applications are vast: medical diagnosis, surveillance, fault detection, condition monitoring, digital twins, etc. Uncertainty, which is a fundamental component of measurements including sensing, must be quantified for risk management, better decision making, and assuring the user that the system is trustworthy. In this tutorial, we show how ML is used for indirect measurement, and how to quantify the uncertainty of ML-assisted measurements to design more reliable and practical I&M systems. We cover uncertainty quantification for both ML regression and ML classification. Finally, we go over a few specific examples, both from existing literature and hands on exercises. Did you know that the IEEE Instrumentation and Measurement Society offers a variety of e-classes? Many of our e-classes allow you to earn CEUs and PHD credits. The e-class for this video tutorial can be accessed here: https://iln.ieee.org/Public/ContentDe... The full list of available videos can be accessed here: https://loom.ly/fikwaQM Keywords: Machine Learning, Uncertainty, Classification, Regression, Measurement