T. Goblick, P. Harmon, and J. Leivent. A multiprocessor for mixed numeric /symbolic computation. Talk given by MIT Machine Intelligence Group, Oct. 4, 1988.  Home/Search   Document Not in Database   Summary   Related Articles

....processing power and storage capacity for moderately large neural nets. A pipelined architecture with a limited instruction set or the use of fast multipliers is appropriate for the processor nodes[36] A notable effort in this direction is the MX 1 16 being developed at MIT Lincoln Laboratories[22]. This multiprocessor is intended for AI applications such as machine vision, that involve both numeric and symbolic computations. It comprises of 16 digital signal processors, each capable of 10M multiply accumulate operations sec, interconnected through a crossbar to one another and to a host ....

....Runs Rochester BBN[4, 8] connected by connectionist butterfly network;VAX simulator[13] CM 2, 64K single bit processors, 13M 64M BP[5] Thinking Machines hypercube network; NETtalk) Corp. 38, 8] VAX or Symbolics MX 1 16 16 digital signal 120M 50M Projected performance MIT processors; Lincoln Labs[8, 22] Lisp machine X MP 2, Two Cray processors 50M 2M Estimated performance Cray Research[8] with shared memory IPS = Interconnects per second; BP = Back Propagation; CP = Counter Propagation; NEO = Neocognitron; HOP = Hopfield; ART = Adaptive Resonance Sources: DARPA Neural Network Report (July 1988 ....

T. Goblick, P. Harmon, and J. Leivent. A multiprocessor for mixed numeric /symbolic computation. Talk given by MIT Machine Intelligence Group, Oct. 4, 1988.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC