Uncertainty Quantification in Deep Learning
Abstract: Deep neural networks have achieved impressive performance on a wide range of machine learning tasks and are gaining popularity in a variety of disciplines. However, a major limitation of these methods is that they provide predictions without estimates of data and model uncertainty. Uncertainty quantification (UQ) is important to know when to trust the model's prediction. UQ enables decision-makers to understand the level of uncertainty associated with each prediction and make informed decisions based on the risks associated with each outcome. UQ is critical for understanding the inherent variations of the data and prediction limitations due to a lack of training data.
In this talk, I will discuss different techniques for uncertainty quantification in deep learning and elaborate on the deep hyperparameter ensemble method. Using a spatiotemporal graph neural network as an example, I will illustrate how a deep understanding of the data and model uncertainty can identify areas for improvement in data collection and modelling processes and help make more informed decisions.
Bio: Tanwi is an Assistant Computer Science Specialist in the Mathematics and Computer Science Division at Argonne. Prior to this role, she was a Postdoctoral Appointee in the same division. Her current research delves into spatiotemporal modeling, scalable data-efficient deep learning, uncertainty quantification, and large-scale machine learning on high-performance computing systems. Before joining Argonne, Tanwi served as a Senior Data Scientist at General Electric. She earned her Ph.D. in Computer Science from the Indian Institute of Technology in Kharagpur, India.
