Dr. Greg Stitt
Thursday, Sept. 3, 1:00–1:15 pm
“PANDORA: A Parallelizing Approximation-Discovery Framework”
Abstract
This talk presents the PANDORA machine-learning framework that complements existing parallelizing compilers by automatically discovering application- and architecture-specialized approximations of code. PANDORA creates approximations that extract massive amounts of parallelism from inherently sequential code by eliminating loop-carried dependencies—a long-time goal of the compiler research community. Compared to exact parallel baselines, results show speedups ranging from 2.3x to 81x with acceptable error for many usage scenarios.
Dr. Joel Harley
Thursday, Sept. 3, 1:30–1:45 pm
“Physics-Guided Signal Processing and Machine Learning for Smart Sensing, Structures, and Materials”
Abstract
Machine learning has produced impressive results with complex, high-dimensional data across many applications that can leverage an abundance of data. In contrast, machine learning has yet to find widespread use in more industrial applications, such as nondestructive evaluation and infrastructure sustainment, where data is more scarce. In this talk, we demonstrate how physics-guided learning strategies can be used in these applications to complete datasets, isolate important information, and detect anomalous behaviors.