This Article 
 Bibliographic References 
 Add to: 
Accuracy vs. Precision in Digital VLSI Architectures for Signal Processing
April 1998 (vol. 47 no. 4)
pp. 472-477

Abstract—The paper provides a sensitivity analysis to measure the loss in accuracy induced by perturbations affecting acyclic computational flows composed of linear convolutions and nonlinear functions. We do not assume a large number of coefficients or input independence for the convolution module, nor strict requirements on the nonlinear function. The analysis is tailored to digital VLSI implementations where perturbations, associated with data quantization, affect the device inputs, coefficients, internal values, and outputs. The sensitivity analysis can be used to measure the loss in accuracy along the computational chain, to characterize the tolerated perturbations, and to dimension the whole architecture.

[1] The Fermi Group, "A Digital Front-End and Readout Microsystem for Calorimetry at LHC: The FERMI Project," IEEE Trans. Nuclear Science, vol. 40, no. 4, pp. 516-530, Aug. 1993.
[2] W.K. Pratt, Digital Image Processing, John Wiley&Sons, New York, 1978.
[3] J. Hertz, A. Krogh, and R.G. Palmer, Introduction to the Theory of Neural Computation. Addison-Wesley, 1991.
[4] S.G. Mallat,“A theory for multiresolution signal decomposition: The wavelet representation,” IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 11, no. 7, pp. 674-693, 1989.
[5] M. Stevenson, R. Winter, and B. Widrow, "Sensitivity of Feedforward Neural Networks to Weights Errors," IEEE Trans. Neural Networks, vol. 1, no 1, Mar. 1990.
[6] C. Alippi, V. Piuri, and M. Sami, "Sensitivity to Errors in Artificial Neural Networks: A Behavioural Approach," IEEE Trans. Circuits and Systems-I, vol. 42, no 6, June 1995.
[7] S. Piché, "The Selection of Weights Accuracies for Madalines," IEEE Trans. Neural Networks, vol. 6, no. 2, Mar. 1995.
[8] J. Holt and J. Hwang, "Finite Precision Error Analysis of Neural Network Hardware Implementations," IEEE Trans. Computers, vol. 42, no. 3, Mar. 1993.
[9] G. Duundar and K. Rose, "The Efects of Quantization on Multilayer Neural Networks," IEEE Trans. Neural Networks, vol. 6, no. 6, Nov. 1995.
[10] B. Hassibi and D.G. Stork, "Second Order Derivative for Network Pruning: Optimal Brain Surgeon," Proc. NIPS5, 1993.
[11] C. Alippi and L. Briozzo, "Accuracy vs. Precision in Digital VLSI Architectures for Signal Processing," Internal Report CNR-CESTIA 97-02, 1997.
[12] C. Alippi and G. Storti-Gajani, "Simple Approximation of Sigmoidal Function: Realistic Design of Digital Neural Networks Capable of Learning," Proc. IEEE-ISCAS,Singapore, June11-14 1991.
[13] M. Valle, D. Caviglia, M. Cornero, G. Nateri, and L. Briozzo, "A VHDL Based Design Methodology the Design Experience of a High Performance ASIC Chip," Proc. Euro—VHDL Conf.,Grenoble, France, Sept.19-23 1994.

Index Terms:
Finite precision representation, neural networks, NSR, sensitivity analysis.
Cesare Alippi, Luciano Briozzo, "Accuracy vs. Precision in Digital VLSI Architectures for Signal Processing," IEEE Transactions on Computers, vol. 47, no. 4, pp. 472-477, April 1998, doi:10.1109/12.675715
Usage of this product signifies your acceptance of the Terms of Use.