Systolic Designs for State Space Models: Kalman Filtering and Neural Networks

  • S.-Y. Kung
  • J. N. Huang


In this paper, a systematic mapping methodology is introduced for deriving systolic and wavefront arrays from regular computational algorithms [10]. It consists of three stages of mapping design: (data) dependence graph (DG) design, signal flow graph (SFG) design, and array processor design. This methodology allows systolic design with many desirable properties, such as local communication and fastest pipelining rates, etc. Based on this methodology, we shall develop systolic array designs for two important applications of adaptive state-space models. One is for the Kalman filtering algorithm which is popular in many digital signal processing applications. The other one is the Hopfield model for artificial neural networks (ANN), which has recently received increasing attention from AI and parallel processing research community.


Kalman Filter Dependence Graph State Space Model Systolic Array Array Processor 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. [1]
    D. H. Ackley, G. E. Hinton, and T. J. Sejnowski. A learning algorithm for Boltzmann machines. Cognitive Science, Vol. 9: pp. 147–169, 1985.CrossRefGoogle Scholar
  2. [2]
    N. H. Farhat, D. Psaltis, A. Prata and E. Paek. Optical implementation of the Hopfield model. Applied Optics, Vol. 24: pp. 1469–1475, May 1985.CrossRefGoogle Scholar
  3. [3]
    G.H. Golub and C. F. Van Loan. Matrix Computations. Johns Hopkins University Press, 1983.Google Scholar
  4. [4]
    J. J. Hopfield. Neural network and physical systems with emergent collective computational abilities. In Proc. Natl.. Acad. Sci. USA, Vol. 79, pp. 2554–2558, 1982.Google Scholar
  5. [5]
    J. J. Hopfield. Neurons with graded response have collective computational properties like those of two-state neurons. In Proc. Natl.. Acad. Sci. USA,Vol. 81, pp. 3088–3092, 1984.Google Scholar
  6. [6]
    J. J. Hopfield and D. W. Tank. Neural computation of decision in optimization problems. Biological Cybernetics, Vol. 52, pp. 141–152, 1985.Google Scholar
  7. [7]
    H. T. Kung and W. M. Gentleman. Matrix triangularization by systolic arrays. Proc. SPIE, Real Time Signal Processing, 1983.Google Scholar
  8. [8]
    R. E. Kalman. A new approach to linear filtering and prediction problems. J. Basic Engineering, 82: pp 35–45, 1960.CrossRefGoogle Scholar
  9. [9]
    C. Koch, J. Marroquin and A. Yuille. Analog “neuronal” networks in early vision. Proc. of National Academy Science, Vol. 83: pp. 4263–4267, 1986.CrossRefGoogle Scholar
  10. [10]
    S. Y. Kung. VLSI Array Processors. Prentice Hall Inc. N. J., 1987.Google Scholar
  11. [11]
    S.Y. Kung. On supercomputing with systolic/wavefront array processors. Proceedings of the IEEE, 72: pp. 867–884, July:1984.Google Scholar
  12. [12]
    S. Y. Kung, S. C. Lo, and P. S. Lewis. Timing analysis and optimization of VLSI data flow arrays. In Proc. IEEE ICPP’86, pp. 600–607, August 1986.Google Scholar
  13. [13]
    S. Y. Kung, J. N. Hwang, and S. C. Lo. Mapping digital signal processing algorithms onto VLSI systolic/wavefront arrays. In Proc. 12th Annual Asilomar Conf. on Signals, Systems and Computers, pp. 6–12, November 1986.Google Scholar
  14. [14]
    S. Y. Kung and J. N. Hwang. Systolic array designs for Kalman filtering. Submitted to IEEE Trans. on Acoustics, Speech, and Signal Processing, 1987.Google Scholar
  15. [15]
    H. Mada. Architecture for optical computing using holographic associative memories. Applied Optics, Vol. 24: 1985.Google Scholar
  16. [16]
    J.G. McWhirter. Recursive least-squares minimization using a systolic array. In SPIE, In Proc. Real Time Signal Processing VI, pp 105–110, SPIE, 1983.Google Scholar
  17. [17]
    M. J. Chen and K. Yao. On realization and implementation of Kalman filtering by systolic array. In Proc. 21st Conf. on Inf. Science and Systems, The John Hopkins University, pp. 375–380, 1987.Google Scholar
  18. [18]
    C. C. Paige and M. A. Saunders. Least squares estimation of discrete linear dynamic systems using orthogonal transformation. SIAM J. Numer. Anal., 14: pp. 180–193, 1977.CrossRefGoogle Scholar
  19. [19]
    M. Takeda and J.W. Goodman. Neural networks for computation: number representations and programming complexity. Applied Optics, Vol. 25: pp. 3033–3046, September 1986.CrossRefGoogle Scholar

Copyright information

© Plenum Press, New York 1988

Authors and Affiliations

  • S.-Y. Kung
  • J. N. Huang

There are no affiliations available

Personalised recommendations