According to space-based theory, visual attention is limited to a local region in space called the attentional field. Visual information within the attentional field is enhanced for further processing while information outside is suppressed. There is evidence that enhancement and suppression are achieved with dynamic weighting of network activity. This paper discusses a neural network that generates the appropriate weights, called the attentional spotlight, given the size and the position of the intended attentional field. The network has three layers. A shunting feedback network serves as the output layer and performs a critical task which cannot be accomplished by feedforward networks.

PDF, PS

In Proceedings of the International Joint Conference on Neural Networks (Singapore), AI91-165, pp. 436-441 1991. Piscataway, NJ: IEEE.

@TechReport{leow:spotlighttr, title={A Neural Network For Attentional Spotlight}, author={Wee Kheng Leow and Risto Miikkulainen}, booktitle={Proceedings of the International Joint Conference on Neural Networks (Singapore)}, number={AI91-165}, institution={Department of Computer Sciences, The University of Texas at Austin}, publisher={Piscataway, NJ: IEEE}, key={IJCNN}, pages={436-441}, url="http://www.cs.utexas.edu/users/ai-lab?leow:ai91-165", year={1991} }