Applications of random-pulse machine concept to neural network design

Neural networks can reach their true potential only when they are implemented in hardware as massively parallel processors. This paper presents the random-pulse machine concept and shows how it can be used for the modular design of neural networks. Random-pulse machines deal with analog variables represented by the mean rate of random-pulse streams and use simple digital technology to perform arithmetic and logic operations. This concept presents a good tradeoff between the electronic circuit complexity and the computational accuracy. The resulting neural network architecture has a high packing density and is well suited for very large-scale integration (VLSI). Simulation results illustrate the performance of the basic elements of a random-pulse neuron.

[1]  Sergio Telles Ribeiro,et al.  Random-Pulse Machines , 1967, IEEE Trans. Electron. Comput..

[2]  Francis Castanie,et al.  Linear mean transfer random quantization , 1984 .

[3]  E. Petriu,et al.  Influence of reference domain unstability upon the precision of random reference quantizer with uniformly distributed auxiliary source , 1983 .

[4]  F. Castanie,et al.  Signal processing by random reference quantizing , 1979 .

[5]  Alan F. Murray,et al.  Integrated pulse stream neural networks: results, issues, and pointers , 1992, IEEE Trans. Neural Networks.

[6]  M. F. Wagdy,et al.  Validity of uniform quantization error model for sinusoidal signals without and with dither , 1989 .

[7]  A. J. Miller,et al.  Moving-average output interface for digital stochastic computers , 1974 .

[8]  P. Mars,et al.  Optimal estimation of digital stochastic sequences , 1977 .

[9]  L. Schuchman.  Dither Signals and Their Effect on Quantization Noise , 1964 .

[10]  Tarun Khanna,et al.  Foundations of neural networks , 1990 .

[11]  John Shawe-Taylor,et al.  Emergent activation functions from a stochastic bit-stream neuron , 1994 .

[12]  M. Wagdy,et al.  Linearizing average transfer characteristics of ideal ADC's via analog and digital dither , 1994 .

[13]  Alfred Fettweis,et al.  A new method to compute correlation functions (Abstr.) , 1962, IRE Trans. Inf. Theory.

[14]  Ke-Yen Chang,et al.  Modified digital correlator and its estimation errors , 1970, IEEE Trans. Inf. Theory.

[15]  Kimmo Kaski,et al.  Pulse-density modulation technique in VLSI implementations of neural network algorithms , 1990 .

[16]  Dario Petri,et al.  Effect of additive dither on the resolution of ideal quantizers , 1994 .

[17]  J. von Neumann,et al.  Probabilistic Logic and the Synthesis of Reliable Organisms from Unreliable Components , 1956 .

[18]  M. Wagdy.  Effect of various dither forms on quantization errors of ideal A/D converters , 1989 .