Article

Article title IMPLEMENTATIONS OF NON-STANDARD MODELS OF NEURON USING NEUROMATRIX NEUROPROCESSOR
Authors O.S. Mezentseva, D.V. Mezentsev, N.A. Lagunov, N.S. Savchenko
Section SECTION V. HIGH-EFFICIENCY COMPUTING ALGORITHMS
Month, Year 06, 2012 @en
Index UDC 004.032.26:004.94
DOI
Abstract Algorithms and software implementations of non-standard neuron models on a hardware platform NeuroMatrix, taking into account the features of its architecture, for the influence researches of the neuron model to the quality of the neural network functioning are given in this paper. Operative and shady matrixes, vector arithmetic logic unit, mask operation, hardware implemented activation functions and neuroprocessor registers-containers effectively used in the developed software modules that implement Pade-neuron, neuron with quadratic adder and neuron with a counter coincidences computations.

Download PDF

Keywords Modeling; neural network system; NeuroMatrix; Pade-neuron; neuron with coincidence counter; neuron with quadratic adder.
References 1. Комарцова Л. Г. Нейрокомпьютеры. – М.: Изд-во МГТУ им. Н.Э. Баумана, 2004. – 41 с.
2. Миркес Е. М. Нейроинформатика: Учебное пособие. – Красноярск: ИПЦ КГТУ, 2002. - 347 с.

Comments are closed.