Multidisciplinary Studies of Integrated Neural Network Systems
Abstract
This project was a joint effort of the David Sarnoff Research Center (Sarnoff), Princeton University, and Robicon Systems, all of Princeton, NJ. It consisted of three sub-projects, each concerned with a similar kind of research - the development of artificial adaptive systems with capabilities similar to those of their biological counterparts. Recent work on neural networks has demonstrated their potential for solving difficult problems in simplified, controlled environments. The next stage in the development of neural networks is their extension to the scale, complexity, and variability of real-world situations. This will not be a simple evolution of existing neural net designs, because it requires the integration of complex adaptive systems whose components have widely differing functions. Fortunately, biological organisms present existing solutions to this problem and neuroscience can now probe in detail the relevant structures. Biological systems are highly adaptive and operate well in extremely complex and variable environments. They accomplish this by partitioning the system into functional sub-units in a quasi-hierarchical structure of neural network modules. We studied three specific examples of this system integration strategy and modeled their operation for the purpose of creating new neural network architectures and control schemes. Neural networks, Auditory localization, Sensor fusion, Neuroscience, Target detection, Motion analysis, Visual cortex, Barn owl, Robotics, Expert systems, Hierarchical architectures, Adaptive control.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 1994
- Accession Number
- ADA278774
Entities
People
- Clay Spence
- David Handelman
- Jack Gelfand
- Jeffrey Lubin
- John Pearson
- Mike Littman
- Steven Lane
- Williams E. Sullivan
Organizations
- Sarnoff Corporation