A single 1.5-V digital chip for a 10/sup 6/-synapse neural network

A single-chip architecture for a 10/sup 6/-synapse neural network is proposed. It operates on a 1.5-V dry cell so that it can be used in portable equipment. An on-chip DRAM cell array stores 10/sup 6/ 8-b synapse weights digitally. These are readily programmable and refreshable. A pitch-matched interconnection and a combinational unit circuit used for summing product provide a tight layout by eliminating bus lines between the memory cell array and the product-summation circuit. A dynamic data transfer circuit directly coupled to the DRAM cell array, together with 1.5-V operation of the entire chip, reduces the power dissipation to 75 mW. The 256 parallel circuits for summing product provide a processing speed of 1.37 gigaconnections/s. The memory and the processing circuits can be integrated on a 15.4-mm*18.6-mm chip by using a 0.5- mu m CMOS design rule.<<ETX>>

[1]  Kazuo Yano,et al.  A 3.8-ns CMOS 16*16-b multiplier using complementary pass-transistor logic , 1990 .

[2]  D. Hammerstrom,et al.  A VLSI architecture for high-performance, low-cost, on-chip learning , 1990, 1990 IJCNN International Joint Conference on Neural Networks.

[3]  Koichiro Mashiko,et al.  A 336-neuron, 28 K-synapse, self-learning neural network chip with branch-neuron-unit architecture , 1991 .

[4]  Katsunari Shibata,et al.  A self-learning neural network composed of 1152 digital neurons in wafer-scale LSIs , 1991, [Proceedings] 1991 IEEE International Joint Conference on Neural Networks.

[5]  Y. Tamura,et al.  A BiCMOS analog neural network with dynamically updated weights , 1992, 1990 37th IEEE International Conference on Solid-State Circuits.

[6]  S. Tam,et al.  An electrically trainable artificial neural network (ETANN) with 10240 'floating gate' synapses , 1990, International 1989 Joint Conference on Neural Networks.

[7]  R. Pinkham,et al.  An 11-million Transistor Neural Network Execution Engine , 1991, 1991 IEEE International Solid-State Circuits Conference. Digest of Technical Papers.

[8]  H.P. Graf,et al.  A reconfigurable CMOS neural network , 1990, 1990 37th IEEE International Conference on Solid-State Circuits.

[9]  Ulrich Ramacher,et al.  Synapse-X: a general-purpose neurocomputer architecture , 1991, [Proceedings] 1991 IEEE International Joint Conference on Neural Networks.

[10]  Bruce A. Wooley,et al.  A Two's Complement Parallel Array Multiplication Algorithm , 1973, IEEE Transactions on Computers.