A New Approach in Cooperative Decision Making in Multi-agent Systems Inspired by Human Visual Cortex

Abstract

In this paper, a new approach in decision making process inspired by human visual cortex has been proposed. In this approach knowledge of a group of agents (training data) will be used for decision-making. The proposed approach tries to meet two fundamental features, i.e., robustness and specificity. The hierarchical model that has been represented in this work, tries to extract the knowledge about the behavior of the system from the training data set by finding the similar training data points. In this model the behavior of the system is governed by the clusters of training data points that in fact every cluster act as an expert. For every new data point, these experts try to predict the label of the corresponding data point and the result of the system is the aggregation of the predictions of different experts. This hierarchical model has been designed inspired by a computational model of object recognition in human cortex. The approach has been used to forecast Mackey-Glass time series and has shown acceptable results.