학술논문

A neural network for attentional spotlight
Document Type
Conference
Source
[Proceedings] 1991 IEEE International Joint Conference on Neural Networks Neural Networks, 1991. 1991 IEEE International Joint Conference on. :436-441 vol.1 1991
Subject
Computing and Processing
Components, Circuits, Devices and Systems
Signal Processing and Analysis
Neural networks
Encoding
Space technology
Humans
Visual system
Psychology
Neurons
Biological information theory
Hardware
Computational modeling
Language
Abstract
The architecture and dynamics of a neural network that generates an attentional spotlight are discussed. The network has three layers. The desired position of the spotlight is indicated by activating one of the input units. The activity of the unit represents the desired size of the spotlight. A shunting feedback network serves as the output layer and produces a spotlight of the desired shape. The spotlight has a fixed maximum size (determined by network parameters) which corresponds to the largest amount of information that a vision system can process in parallel. The spotlight generation network was simulated numerically using a forward Euler method.ETX