Region Finding for Attention Control in Consideration of Subgoals

Neural Network World. International Journal on Neural and Mass-Parallel Computing and Information Systems.
Special issue on Neurofuzzy'96 - IEEE European Workshop, 6 (1996) no.3, pp.305-313.

Fred Hamker & Horst-Michael Groß

The basic idea of our approach is to avoid the separation of perception and generation of action by fusing both parts into one neural process, an attentive vision process. This means, information extracted form the environment is selected according to the relevance of the systems intended action. A neural network for the selection of task relevant visual regions is introduced. It consists of interacting columns with local excitatory and global inhibitory coupled feedback. The basic idea is that lateral cooperation is used as a way to integrate subgoals. Possible subgoals are the size of regions, the security of a local feature hypothesis and the valuation of the hypothesis for the task. The input activity of different feature maps evokes several hypothesis-assemblies, which compete in a recurrent cycle. Thus one assembly of cooperating neurons remains active, others are suppressed with regard to the relevant subgoals. Three simplified simulations show the performance and indicate a segmentation approach.

Zurück zu Fred Hamker