Seminar: Learning from imbalanced and drifting data streams
Alberto Cano Rojas
Associate Professor, Department of Computer Science
Virginia Commonwealth University
Friday, September 13, 2024
9:00 - 10:00AM
1100 Torgersen Hall
Abstract
Data streams are ordered and potentially unbounded sequences of instances that arrive continuously with time-varying intensity, speed, and characteristics. Class imbalance and concept drift pose significant challenges to adapt classifiers to the evolving nature of data considering non-stationary properties. This seminar will present the many challenges and potential solutions to the difficulties of learning robust online classifiers from imbalanced and drifting streaming data. We will focus on ensemble-based approaches that self-adapt to dynamic changes in the data distributions over time. We will discuss performance metrics and benchmarks to present a comprehensive and reproducible evaluation of classifiers in the state of the art.
Biography
Alberto Cano is an Associate Professor with the Department of Computer Science, Virginia Commonwealth University (VCU), where he heads the High-Performance Data Mining lab. He is also the Faculty Director of the VCU High Performance Research Computing facility, responsible for university-wide research computing. He obtained his PhD degree in Computer Science from the University of Granada, Spain, in 2014. His research is focused on machine learning, data streams, high-performance computing on GPUs, and evolutionary computation. His research has been funded by NSF, State of Virginia, Amazon, Hamilton Beach, CCI, VCU, and Korea. He has published over 59 articles in high-impact factor journals and 58 conference contributions. Dr. Cano serves as Area Editor of the journal Information Fusion (IF 14.7). He is recognized among the top 2% most cited researchers in Artificial Intelligence by Stanford University ranking.