An improved hebbian neural network with dynamic neuronal life and relations

In the last years, artificial intelligence has been an important field as the environments in which human-made devices have to operate become more and more complex, and designing a new algorithm for each environment can be very time and resources consuming. Neural networks have been successful in a lot of applications, since the same basic implementation can be used in a virtually unlimited number of situations, by modifying the structure of the network and the tests that are used for constructing it. In this paper we present an improved hebbian neural network that has the capability of adding new neurons to it and can connect neurons using an association rule. Since the main problem in neural network design is the actual construction of the inter-neuronal relations, we try to solve this issue at least partially by allowing the network to modify itself depending on its response to different stimuli.