Neural networks-their use for the routing of integrated circuits

The implementation of a channel router using a number of pretrained neural networks is described. These networks are interconnected to construct a complete routing system. The router is implemented as a software simulation. Results show that it is possible to construct a complex neural system using smaller neural networks. The resultant channel router provides a definition in a form suitable for implementation as a parallel neural system. Although many features of neural networks, such as in service learning, are not used, the system could provide high-speed routing, either by implementing each artificial neuron in hardware or with an appropriate parallel hardware accelerator