You are here
Theoretical and Applied Data Science Lunch-n-learn - Ardhendu Tripathy
Presenter: Ardhendu Tripathy
Title: Adaptive Algorithms in Machine Learning
Abstract: Modern Machine Learning has resulted in several successes, from developing the best players in games like chess and go, to health monitoring that can predict the onset of illness. This talk will explore the role of active sampling, which allows algorithms to solve problems more efficiently by collecting only the necessary data. I will start by presenting a new algorithm to cluster random variables based on their means. For example, we can apply our framework to identify "safe" neighborhoods in Chicago using 7 times fewer queries than passive learning. Next, I will describe an algorithm that uses noisy estimates of distances to identify nearest neighbors. In a real-world dataset consisting of human responses, our approach correctly identified twice as many nearest neighbors compared to a state-of-the-art passive approach. I will conclude by briefly outlining an approach to obscure sensitive information in a dataset without notably affecting its utility. The algorithms and frameworks described in this talk can have many applications in areas such as reinforcement learning and explainable machine learning.
Bio: Ardhendu Tripathy is a postdoctoral associate at the University of Wisconsin-Madison. His current research focusses on designing and analyzing algorithms and methods in statistical learning theory. He obtained his Ph.D. in Electrical and Computer Engineering from Iowa State University, working with Aditya Ramamoorthy on problems in computing over networks. In the past he has done internships at Mitsubishi Electric Research Labs in Cambridge, MA and Qualcomm India.
After the presentation, there will be a short time for discussion and questions afterwards.