Title :
Implementation of a neuron dedicated to Kohonen maps with learning capabilities
Author :
Hochet, B. ; Peiris, V. ; Corbaz, G. ; Declercq, M.
Author_Institution :
Swiss Fed. Inst. of Technol., Lausanne, Switzerland
Abstract :
A compact implementation of Kohonen-oriented neurons with learning capability, using both analog and digital techniques, with standard low-cost CMOS technologies, is described. Each synaptic weight is stored as a discrete voltage on a capacitor. Leakage currents are compensated by a special circuitry, which is also used for learning. Input signals are frequency-coded pulse streams, and the synaptic multipliers are reduced to simple AND gates
Keywords :
CMOS integrated circuits; learning systems; logic gates; neural nets; AND gates; Kohonen maps; Kohonen-oriented neurons; analogue techniques; digital techniques; discrete voltage; frequency-coded pulse streams; learning capabilities; low-cost CMOS technologies; synaptic multipliers; synaptic weight; CMOS technology; Capacitors; Circuits; Costs; Frequency; Leakage current; Neurons; Self organizing feature maps; Vectors; Voltage;
Conference_Titel :
Custom Integrated Circuits Conference, 1990., Proceedings of the IEEE 1990
Conference_Location :
Boston, MA
DOI :
10.1109/CICC.1990.124802