The DAN Research Group

Home About DANs Projects Past Activity Team Members Contact

Welcome to the home of the DAN Research Group, where we study dynamic assocative networks and conceptual implications relating to them.

This research began in the 1990's with work by Anthony Beavers and Hiten Sonpal on quality-control mechanisms in early search engines. These mechanisms functioned by limiting the scope of a search prior to running it, thereby making search return sets both relevant and reliable. At the algorithmic level, the process of limiting the scope of a search was a matter of properly gating information as it flowed over a partially-connected directed graph. In the years following, this idea of gating information flow over a selected portion of a directed graph to determine relevance was adapted by Beavers into a method for building content-addresssable memory without the need for the exogenous training techniques used in today's impressive AIs. Though the models built here do nothing as impressive (yet), what can and cannot be achieved by modeling in this localist manner has not yet been fully worked out. Given this fact, initial success with small models thus far, and the great deal at stake if this research is on the right track, it seems both imperative and timely to push it as far as we can. To learn more, see About DANs.

Early funding for this research was provided by the National Endowment for the Humanities and the University of Evansville.

Anthony F. Beavers, PhD, Director

afbeavers@afbeavers.net

Report abuse