Abstract
A general-purpose Neural Signal Processor MA16 is presented the architecure of which is guided by an analysis of today's neural algorithms. The MA16 executes the elementary algorithmic strings which are compute-bound and shared by all neural nets; operations which are not time consuming are let to hardware off-the-shelf. Digital design is chosen because of flexibility and computation accuracy. The throughput is 800 million connections per second at a clock frequency of 50 MHz (1 connection = 16 bit). This performance is valid for arbitrary networks provided they consist of more than 16 neurons each comprising more than 16 synapses. Depending on the needs of the application under consideration a linear or 2-dimensional array of VLSI chips can be constructed in order to provide sufficient processing power and weight memory.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
References
DARPA Neural Network Study, pp. 34 (figure 2.14–15), pp. 330 (figure 28.5), AFCEA.
H.P. Graf, et al., “Recent developments of electronic neural nets in USA and Canada,” U. Ramacher, U. Rückert, and J. Nossek, eds., inProceedings of the 2nd International Conference on Microelectronics for Neural Networks, 1991, pp. 471–490.
Y. Hirai, “Hardware implementation of neural networks in Japan,” pp. 435–454, ibid.
M. Griffin, et al., “An 11-million transistor neural network execution engine,”Digest of Technical Papers of the Int. Solid State Circuits Conf, vol. 33, 1991, pp. 180.
M. Yasunaga, et al., Design, fabrication and evaluation of a 5-inch wafer scale neural network LSI composed of 576 digital neurons,”Proceedings of the IJCNN-90, vol. II, 1990, pp. 527–535.
L. Curran, “Wafer scale integration arrives in ‘disk’ form,” inElectronic Design, Oct. 26, 1989, pp. 51–54.
U. Ramacher, “Hardware concepts of neural networks,” in R. Eckmiller,Advanced Neurocomputers, Elsevier, 1990, pp. 209–218.
U. Ramacher, “Hamiltonian dynamics of neural networks,”Proceedings of the IJCNN-91, 1991.
L.B. Almeida,IEEE Int. Conf. on Neural Networks, 1987, pp. II-609.
F.J. Pineda,Phys. Rev. Letters, vol. 59, 1987, pp. 2229.
B.A. Pearlmutter,Neural Computation, vol. 1, 1989, pp. 263.
U. Ramacher et al., “Design of a 1st generation neurocomputer,” in U. Ramacher, U. Rückert, eds.VLSI Design of Neural Networks, Boston: Kluwer Academic Publishers, 1991.
U. Ramacher and M. Wesseling, “WSI architecture of a neurocomputer module,”Proc. of the IEEE Int. Conf On Wafer Scale Integration, 1990, pp. 125–130.
U. Totzeck, F. Matthiesen, S. Wohlleben, and G. Noll, “CMOS VLSI implementation of the 2D-DCT with linear processor arrays,”Proc. of IEEE ICASSP-90, pp. 937–940.
Author information
Authors and Affiliations
Rights and permissions
About this article
Cite this article
Ramacher, U., Beichter, J. & Brüls, N. A general-purpose signal processor architecture for neurocomputing and preprocessing applications. J VLSI Sign Process Syst Sign Image Video Technol 6, 45–56 (1993). https://doi.org/10.1007/BF01581958
Received:
Revised:
Published:
Issue Date:
DOI: https://doi.org/10.1007/BF01581958