Title :
Scale invariance and the upper limits of interconnection complexity
Author :
Christie, Phillip ; Dorricott, Brian
Author_Institution :
Dept. of Electr. Eng., Delaware Univ., Newark, DE, USA
Abstract :
A definition of interconnection complexity in terms of a constant scale-invariant dimensions of information flow is presented. The proposal that all interconnection networks must be characterized by a constant dimensionality if they are to interconnect an arbitrarily large processing system is discussed. This construction hypothesis is applied to neural networks. The results of a series of experiments to determine the optimum dimensionality are reported
Keywords :
circuit layout; fractals; multiprocessor interconnection networks; network topology; neural nets; parallel architectures; algorithms implementation; constant scale-invariant dimensions; fractal description; fractal neural array; information flow; interconnection complexity; interconnection networks; large processing system; neural networks; optimum dimensionality; placement algorithm; upper limits; Cooling; Equations; High performance computing; Logic; Mathematical model; Multiprocessor interconnection networks; Neural networks; Pathology; Power dissipation; Power system interconnection;
Conference_Titel :
Circuits and Systems, 1990., IEEE International Symposium on
Conference_Location :
New Orleans, LA
DOI :
10.1109/ISCAS.1990.111965
