Construction of a Connectionist Network Supercomputer
Technical progress rept. 1 Aug-1 Nov 1992
CALIFORNIA UNIV BERKELEY
Pagination or Media Count:
This quarter was the second quarter of funding by the ONR for the project. With the project fully staffed we turned our attention to the reexamination of the paper design submitted in the proposal. Many new ideas arose as we began to work through the details of the system. To limit the time spent on these investigations, we set a goal of early October for a CNS-1 Architecture Specification. The document would be used as an internal design specification. Also, because we were at the point in the project where we have settled major architectural decisions, we wished to get feedback from other researchers in the field. In the second week of October, we distributed the Architecture Specification to approximately 25 experts in the fields of computer architecture and neurocomputing. With the Architecture Specification written and distributed we held a design review meeting on October 26. The meeting brought together approximately 40 researchers some part of out team, other colleagues from UC Berkeley and ICSI, and some outside experts. The primary focus of the meeting was the hardware architecture of the CNS-1.
- Computer Hardware