Efficient implementation of neural networks in an object oriented programming environment

in Proceedings of Neuro Nimes '91, Fourth International Conference, pp. 733-736, EC2, Paris, 1991.

Abstract	We present an object oriented environment for the design of neural network applications. The environment is based upon an object oriented database management system and includes a programming language suited for neural network simulations. In this language, it is easy to construct and modify complex network structures in a modular fashion. Usually, using an object oriented language would increase execution time considerably, but by using a higher level of data abstraction it is possible to overcome this problem. Some implementational techniques are discussed that will make the execution more efficient. Also, the environment described here is well fitted for the construction of graphical user interfaces. The network topology can be defined by using this interface as well as by using the integrated programming language.

Full paper	Full paper in pdf format.