Communications of the ACM
From Viruses to Social Bots, Researchers nearth Structure of Attacked Networks
Researchers at the University of Southern California are exploring how viruses interact with proteins in the body using a new statistical machine learning technique.
Credit: Shutterstock
University of Southern California (USC) researchers are exploring how viruses interact with proteins in the body using a new statistical machine learning technique.
The protein interaction network simulates each protein as a node, with nodes linked by an "edge" if they interact.
Said USC's Yuankun Xue, "An attack by a virus is analogous to removing certain nodes and links in this network," which means the original network can no longer be observed.
Xue added that the statistical machine learning framework is designed to trace all possible effects of a viral attack, to identify its most likely impact.
The researchers included attack influence and causality, or "adversarial intervention," in the learning algorithm, rather than make it a random sampling procedure; the framework's resulting generality means the technique could be applicable to any network reconstruction challenge entailing adversarial attacks, in fields as diverse as social science, neuroscience, and network security.
From USC Viterbi News
View Full Article
Abstracts Copyright © 2019 SmithBucklin, Washington, DC, USA
No entries found