A Biological neural network (BNN) is composed of a vast number of neurons interconnected by synapses. It has the ability to process information and generate a specific pattern of electrical activity. To analyze its interior structure and exterior properties, computational models were combined with experimental data and one computer simulation system was implemented. As BNN is a complicated nonlinear system and the simulation deals with a great amount of numeric computations,so data representation and computation scheme are critical to simulation process. In this paper, Object-oriented data representation (OODR) was designed to have sharable and reusable properties, and one novel hybrid computation scheme was presented. With OODR, data share and computation share were simultaneously achieved. According to the hybrid computation scheme, individual computation method was applied to corresponding object based on its model characteristics and the computation efficiency was obviously increased. Now they were adopted in one BNN simulation system which was implemented in platform independent language JAVA. As the simulation system took advantage of the data representation and the computation scheme, so its performances were greatly improved, and it has got practical applications in many countries.
展开▼
