[ad_1]
The social network Linkedin will tell a user how he is connected to another. In real life, connection points are not always obvious. However, identifying models or relationships and commonality between entities is an extremely important task for companies, biologists, doctors, patients, etc.
A new computer tool developed in the USC Viterbi School's laboratory, Ming Hsieh, professor in the Electrical and Computer Engineering Department, Paul Bodgan, in collaboration with Professor Ming Hsieh, Edmond Jonckheere, is able to: quickly identify affiliations and hidden interrelationships between groups / elements / people accuracy than existing tools.
The researchers at Bogdan's lab are sort of detectives and the puzzle they're trying to figure out is how one index, person, element or action is linked to another entity. Imagine a laboratory dedicated to a "Six Degrees of …" scientist to discover hidden interrelationships. Researchers studying complex networks call this the "problem of community detection", identify and map individuals or elements that have in common and how they are connected.
Such a computational tool could be exploited by various groups: political strategists trying to find common values or attributes of overlapping voters; or biologists who wish to predict the potential for side effects or interactions of a drug – without having to conduct multi-year experiments. Their research is also being done to identify the parts of the brain that perform the same functions – essential information for neuroscientists and people with brain damage, to anticipate whether certain areas of the brain might support tissue function. harmed. One can also imagine that the algorithm of this lab is looking for contact points for seemingly unrelated information.
Methodology / Proof of concept
PhD student Jayson Sia who worked on the research indicates that the algorithm he developed, the Ollivier-Ricci curvature-based community identification (ORC), was tested and validated on four sets of data from the real world known, the field for which the goal is to find. the point of connection between the "nodes" or the individual / individual elements of a group by examining the links that unite them or what is called in the technical jargon the "edges". Datasets include a Drug Interaction Network, the Zachary Karate Club; an affiliation to a university football conference; and a set of more than 1000 political blogs.
The main author, Sia, explains: "In this article, we used a new geometric approach via the curvature of Ollivier-Ricci, which provides a natural method for discovering the inherent structures of the network community. "
The curvature in the geometric context, Sia explains, "essentially measures the surface difference (or" curves "of a surface) .The geometry of surfaces is related to the study of map projections and to the extent Distances in a Curved Surface The curvature of Ollivier-Ricci extends this concept of "curvature" to networks whose edges, with positive curves, are "well connected" and naturally form a "community." Negatively curved edges on the other hand, are interpreted as "bridges" between communities, and cutting such edges would isolate the flow of information between communities. "
Source of the story:
Material provided by University of Southern California. Note: Content can be changed for style and length.
Source link
