Decision making using Topological Information

摘要

This paper proposes a classification method using neural network that leams topological information. With this method, the system can make relevant decision even though it has not sufficient information. The neural network classifies the input data by comparing with already learned topologies. This process consists of 2 phases. First, the system collects precise data and learns the topological information of given data. Then, the neural network classifies the data into "pre-learned" topologies. The simulated World Cup Robot Soccer (RoboCup) is used as problem domain to evaluate the proposed system.