An integrated neurocomputing system

The design of a comprehensive neurocomputing system is discussed. This system comprises two distinct neural network environments: programming and execution. The programming environment is Pygmalion. An execution environment is proposed that extends Pygmalion by providing a range of target hardware, from conventional machines to parallel neurocomputers and specialized chips generated by silicon compilation. The authors present their view of the neurocomputing system in three stages. First, they describe the Pygmalion programming environment, then the integrated system, and finally the neurocomputer architecture. They discuss the design of neural chips for the neurocomputer, introducing their basic architectures and presenting the system´s silicon compiler