On neural network programming
In this paper we investigate how a neural network can be regarded as a massive parallel computer architecture. To this end, we focus our attention not on a stochastic asynchronous model, but on the McCulloch and Pitts network [MP43], as modified by Caianiello [Cai61], and we specify what we mean for the environment in which the network operates: it is essentially the entity assigning meaning to the network input and output nodes. Changing the environment definition implies dealing with different neural architectures.
To show how to program a neural architecture, we introduce a model of environment helping us in choosing functions suitable to be pipelined, as in a data flow architecture. As an example, we sketch the working of a parallel multiplier function.
KeywordsNeural network Parallel processing Parallel architectures Scientific topic Architectures languages and environments
Unable to display preview. Download preview PDF.
- [Cai61]E. R. Caianiello. Outline of a theory of thought processes and thinking machines. J. of Theor. Biol., (2):204–235, 1961.Google Scholar
- [Den80]J. B. Dennis. Data flow supercomputers. Computer, 48–56, november 1980.Google Scholar
- [DP90]P. De Pinto. The implementation of the assembler associated to a general purpose neural network. In Third Italian Workshop on “Parallel Architectures and Neural Networks”, Vietri sul Mare (SA), May 1990.Google Scholar
- [DPLS90]P. De Pinto, F. E. Lauria, and M. Sette. On the Hebb rule and unlimited precision arithmetic in a MacCulloch and Pitts network. In R. Eckmiller, editor, Advanced neural computers, pages 121–128, International Symposium on: Neural Networks for Sensory and Motor Systems, Elsevier Sci., Amsterdam, 1990.Google Scholar
- [Hew85]C. Hewitt. The challenge of open systems. BYTE, 223–242, April 1985.Google Scholar
- [HU79]J. E. Hopcroft and J. D. Ullman. Introduction to automata theory, languages and computation. Addison-Wesley Pub. Co., 1979.Google Scholar
- [Lau88]F. E. Lauria. A connectionist approach to knowledge acquisition. Cybernetics and Systems: an Intern. Journ., (5):19–35, 1988.Google Scholar
- [Lau89]F. E. Lauria. A general purpose neural network as a new computing paradigm. In E. R. Caianiello, editor, Second Italian Workshop on “Parallel Architectures and Neural Networks”, pages 131–144, World Scientific Pub., Singapore, 1989.Google Scholar
- [MP43]W. S. McCulloch and W. Pitts. A logical calculus of the ideas immanent in the nervous activity. Bull. Math. Bioph., 115–143, 1943.Google Scholar
- [RM86]D. E. Rumelhart and J. L. McClelland. Parallel distributed processing. Volume 1, MIT Press, Cambridge, MA, 1986.Google Scholar
- [Set90]M. Sette. An OCCAM simulation of a general purpose neural network. In Third Italian Workshop on “Parallel Architectures and Neural Networks”, Vietri sul Mare (SA), May 1990.Google Scholar
- [Tra91]G. Trautteur. Problems with Symbols. A commentary to: Herbert Simon, “Scientific Discovery as Problem Solving”. RISESST, (1), 1991. To be published.Google Scholar