Author ORCID Identifier

Ayorinde Emmanuel Olatunde

Document Type

Poster

Publication Date

Summer 8-6-2024

Abstract

As the use of Machine learning models in science and engineering continues to increase, there is an increasing need for quantifying the uncertainties inherent in the predictions of these models. The more complex a model is, the more the uncertainties in its predictions increase. Amongst the plethora of methodologies used in quantifying uncertainties lies Gaussian Process Regression (GPR). GPR surmounts some of the popular shortfalls of other state-of-the-art methodologies. Although GPR has some quick wins in its application for uncertainty quantification, it is plagued with some shortfalls, such as scalability issues when the feature space increases as well as an increase in computational time. Our current study compares the computational time besides quantifying the uncertainties in the predictions from the machine learning models across different covariance structures. Specifically, we used 2D diffraction patterns recorded on a 2D area detector using high-energy X-ray diffraction (HEXRD) to predict the volume fraction of the β-phase of Ti-6Al-4v (Ti64). To achieve this, we reduced the features through Principal Component Analysis and used components that account for 95%, 99% and 99.5% variation in the 2D diffraction images for each of the four datasets used respectively. With the current methodology, we have scaled the application of GPR to high-dimensional cases while we are exploring other methodologies that will reduce the computational time when the sample size becomes large. The goal of the project is to integrate these methodologies to achieve scalability with shorter computational time.

Keywords

Uncertainty Quantification Machine Learning Gaussian Process Regression

Publication Title

10th Annual Data Science in Engineering and Life Sciences Symposium

Rights

© The Author(s)

Creative Commons License

Creative Commons Attribution-NonCommercial 4.0 International License
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.