An adaptive computing system

Several natural systems are comprised of networks of similar cells or processes which possess an intelligence or overall stability when interacting among themselves. Well known though not fully understood examples are the central nervous system or the activity of genes during cell replication. Theoretical studies of network systems have been restricted by complexity to networks of cells possessing linear properties. Computer simulation of networks with non-linear cells is impractical because of the very large number of internal states possessed by networks of non-trivial size. This thesis presents a practical approach to a system capable of modelling some networks using Random Access Memory (R.A.M.) circuits. Interconnections may be arbitrarily specified between the address inputs and data outputs of R.A.M. elements and external network data inputs and outputs. The subsequent out­put state-activity of the network may be "adapted" according to external data using the R.A.M. write enable inputs. A custom made R.A.M. integrated circuit is described which has provision for the simultaneous setting and resetting of its store. This is used in a mini-computer controlled network of 1,280 R.A.M.'s with interconnections specified by software.