Unsupervised learning of concept for action planning

The integration of patterns and symbols is an important study in the artificial intelligence and neural networks. Such integration problems often take place in action planning in artificial intelligence. It is difficult to combine the pattern and the symbol directly. The symbols are operated by a sequence of the action to attain the goal object. In this paper, the integration between the symbols and the action sequence was carried out in the neural network. To realise the integration, first the representation of the symbols is realised in the state map representation which is a kind self-organizing feature map. Next, an unsupervised learning algorithm is developed for the knowledge acquisition on the state map representation in the neural network. To clarify these methods developed here, computer simulation is carried out, in the neural network.