When
Tuesday, November 12, 2024 - 2:00 p.m.
Artin Majdi
Postdoctoral Research Associate II
University of Arizona
"Machine Learning for High-Stakes Decision Making: From Theory to Real-World Applications"
ECE 530 | Zoom link
Abstract: The integration of machine learning into high-stakes decision-making processes presents significant challenges in reliability and practical implementation. This seminar will present systematic approaches to addressing these challenges through both theoretical frameworks and validated real-world applications across multiple domains. The presentation will examine a progression of research contributions in machine learning reliability. The discussion will highlight published works such as the development of "Drive-Net," a robust supervised learning system for driver distraction detection and automated medical image analysis, including a novel multi-planar cascaded architecture for thalamic nuclei segmentation that demonstrates superior performance across varying magnetic field strengths and contrasts. Additionally, the seminar will present contributions to environmental monitoring through machine learning-based prediction of stratocumulus cloud clearings. The seminar will also present two patented innovations: a thermal imaging-based system for non-invasive autonomic nervous system monitoring and a multi-sensor platform for respiratory pattern analysis in COVID-19 patients. These developments showcase the successful translation of theoretical advances into practical solutions. The presentation will further present methodological advances in uncertainty quantification, introducing novel frameworks for taxonomic-aware classification and consensus-based crowdsourcing that address fundamental challenges in machine learning reliability. Dr. Majdi’s current research focuses on the development of AI-driven systems for chronic wound care, supported by various federal and industry funding. This work synthesizes multimodal sensing technologies with advanced machine learning approaches through collaborations with leading healthcare institutions and companies, including USC, Medline, Vibrant Health, and Kent Imaging Inc. The presentation will demonstrate how this research program establishes new paradigms for quantitative wound assessment and personalized intervention strategies while advancing the broader field of AI-enabled healthcare.
Biography: Artin Majdi completed his PhD in electrical and computer engineering at the University of Arizona in 2023, building upon dual master's degrees in ECE and biomedical engineering. His research has produced significant contributions to machine learning applications in autonomous systems, medical imaging, and environmental monitoring, documented through peer-reviewed publications in leading journals and U.S. patents. Currently a postdoctoral researcher at the University of Arizona, Dr. Majdi works on integrating sensing technologies with advanced data analysis and machine learning for chronic wound management.