Abstract – In machine learning, bias is defined pedagogically as the distance between the prediction of a value and its true value. Colloquially, bias is an error induced by assumptions, unproven or otherwise. Socially, bias is exhibited prejudice in favor of (or against) a person or a group of people. These three definitions of the same term conflate uniquely in machine learning as decisions made by these models continue to proliferate our daily lives. The decisions not only determine the allocation of network resources and the speed and orientation of autonomous vehicles, they also affect facial recognition, active policing, surveillance, border crossings, hiring, recidivism, health care, college admissions and much more. A fair and bias-free model is an inescapable requirement especially as these models make decisions that impact lives and inform public policy.
This panel of experts will address the quality and reliability of the decisions made by systems that use machine learning models. This panel will discuss the potential to identify bias when it appears in our systems, assist in that process by raising awareness of the reliability and quality of AI and ML models, examine the influence of sources of bias, and identify how to create fair models.
- Michael Tsikerdekis, PhD, Department of Computer Science, Western Washington University
- Pejman Khadivi, PhD, Department of Computer Science, Seattle University
- Prof. Cynthia Hood, Department of Computer Science, Illinois Institute of Technology
- Speaker 4
Abstract – 5G offers two architectures to provide services. Non-Standalone (NSA) anchors the control signaling of 5G to the 4G base station. While the Standalone (SA) scheme has the 5G base station directly connected to the 5G core network and does not depend on the 4G network at all. These paths offer the ability to reuse the 4G infrastructure to support 5G NR (NSA) or build a brand new 5G only infrastructure (SA). This panel will discuss and debate the merits the two paths.
- Speaker 1
- Speaker 2
- Speaker 3