For 26 October 2022
Hi WN@TL Fans,
Reaching back at least 2,500 years, the idea and the ideals of a liberal education—the type of education best suited to keep a free person free—built on a foundation heavy in math and the ability to recognize, enjoy and exploit patterns in Nature.
While in the medieval era the trivium honed skills in grammar, logic and rhetoric, in contrast the quadrivium drew forth the learner’s talents in geometry, astronomy, arithmetic and music.
Here the focus was on developing in humans their ability to learn and to discern, to predict & to project, and most fundamentally, to make judgements and to act upon them.
In the past 50 years or so, humans have sought to teach their machines how to perceive patterns, how to distill data, how to assess and decide. It is not the abacus on our counting table, but rather the cell phone in our pockets, that is the hand-held icon of this new age of epistemology.
On October 26 Ramya Vinayak of the Department of Electrical and Computer Engineering tunes us in to a defining feature of our times: “Learning from Societal Data.”
Description: Machine learning algorithms learn to make predictions and inferences using large quantities of data. More than half a century of advances in this field have led us to build very good systems that identify spam emails, make our phone cameras detect where to focus when taking pictures, and recommend relevant items or movies on e-commerce platforms. We are now using machine learning algorithms to many critical societal applications in health care, finance, criminal justice systems, and governance to aid in decision making which have far reaching consequences on our lives in the present and the future. However, the nature of the data that we learn from and the consequences of making wrong inferences in these applications are very different.
In many societal applications, the data comprises people from diverse backgrounds. The inferences we can draw from such societal-scale datasets are often severely limited not by the number of people in the data but rather by limited observations available for each individual. Therefore, addressing these challenges due to diversity among the population and the limited observations per individual are critical. My research focuses on tackling these limitations both from theoretical and practical perspectives. In this talk, I will provide a high-level overview of how machine learning algorithms learn from data, what are the key challenges to learning from data from diverse people, and some of the approaches being developed in my research group and from other researchers in the field to tackle them.
Bio: Ramya Vinayak is an assistant professor in the Department of Electrical and Computer Engineering and affiliate faculty in the Department of Computer Science at UW-Madison. Her research focuses on machine learning, statistical inference and crowdsourcing.
Prior to joining UW-Madison she was a postdoctoral researcher at the University of Washington in Seattle. She obtained her Ph.D. and masters degrees in Electrical Engineering at Caltech in Pasadena, California. She grew up in the southern part of India and obtained her undergraduate degree in Electrical Engineering at Indian Institute of Technology Madras before starting her academic research journey. In her spare time, she enjoys hiking, cooking and painting.
Machine learning and Optimization Group at UW-Madison: https://mlopt.ece.wisc.edu/
Next week (November 2) Ozioma Onkonkwo of Geriatrics & Gerontology in the School of Medicine & Public Health will give us a look under the hood at “SuperAgers,” people over the age of 80 who have superior memory capacity.
“I am thrilled to bring the SuperAging study to Wisconsin,” Okonkwo said upon the news in December that UW was named one of five new international sites for the study.
“Understanding the factors that enable SuperAgers to remain cognitively sharp has the potential to provide researchers with useful clues about ways to help us all maintain optimal brain health into late life.”
Okonkwo and his team will eventually recruit 100 participants to join the SuperAging study.
Other sites in the new multi-center SuperAging consortium coordinated through Northwestern University are located in Michigan and Georgia, as well as southwest Ontario.
The SuperAging consortium is led by Emily Rogalski, PhD, associate director of the Mesulam Cognitive Neurology and Alzheimer’s Disease Center at Northwestern University Feinberg School of Medicine, Changiz Geula, PhD, research professor at the Mesulam Center, and Marsel Mesulam, MD, director of the Mesulam Center.
On November 9 Minhal Gardezi of Physics takes us back to the 14th Century with her exploration entitled, “Can X-rays Trace the Origins of Printing?”
Description: With the advent of the Gutenberg printing press in the mid 15th century came a boom in literacy, revolutionizing the way Europeans standardized and disseminated information, and establishing the printing press as one of humanity’s most important inventions. While multiple original Gutenberg Bibles have been preserved to the present day, surprisingly little is known about the actual press itself, leaving several unanswered questions about the origins of printing.
However, Gutenberg’s press is only a fraction of the story of early human print. While the first Gutenberg Bibles were being print, thousands of miles away, Korean artisans were building upon hundreds of years of diverse printing experience. The earliest known preserved document printed on a moveable type printing press is a Korean Buddhist text called Jikji, printed in 1377, nearly 80 years before Gutenberg’s Bibles. A wealth of documents proceeding Jikji remain preserved, and their study is critical to understanding early human print.
The questions remain: How were early Eastern and Western printing presses constructed? And how, if at all, were they connected? Here we bring a physicist’s perspective to the investigation. We use synchrotron-generated X-rays to study the makeup of early printed pages from both regions, including leaves of an original Gutenberg Bible and a Korean Confucius text. Collaborating with a large team of scholars from around the globe, we seek to shed new light (literally) on the origins of print.
Bio: Minhal Gardezi is a graduate student in the University of Wisconsin-Madison Department of Physics conducting X-ray research under Dr. Uwe Bergmann. She focuses on using X-ray spectroscopy methods to examine cultural heritage artifacts. Minhal graduated from Wellesley College in 2020 with a degree in Physics and Computer Science. She completed her undergraduate thesis and post-graduate research fellowship in Condensed Matter Physics at Harvard University. She is interested in using her interdisciplinary background to collaborate with scholars outside her field and broaden the direct applications of X-ray Physics. She is interested in science communication and education, seeking new ways to make high-level research and theory accessible to a general audience.