Feasibility of Artificial Attention at Beyond-Human-Scales
Abstract
Data overload is no longer the exception; it is the rule. Unfortunately, there are few instances where an overwhelming amount of data collected is managed successfully, more typically work-arounds such as filtering, ignoring, deleting, or increase manpower are used to continue to make progress toward goals. One exceptional and successful instance of managing data overload from sensing is human visual attention. Coincidentally, advances in technology over the past 30 years have led to computational models of attention that make it possible to simulate attention processes. In the present work we develop and build a computational model of attention, called Artificial Attention, that operates over a network of sensors like those referred to by the United States Air Force as layered sensing systems. The present work differs from previous computational models of attention that only operate on a single sensor with a narrow and fixed fieldof- view by being scalable to operate over networks of sensors with no predefined field-of-view. The computational model of attention is used to examine what conceptual and practical advances are needed to scale computational models of attention to handle the multiple sensor feeds found in layered sensing systems. The results of preliminary testing show that several hidden assumptions behind current computational models of attention block scaling Artificial Attention to layered sensing system scales.
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 01, 2012
- Accession Number
- ADA567742
Entities
People
- Alexander Morison
- David Woods
Organizations
- Ohio State University