Title :
Implementation of artificial neural network with adaptable topology on fixed-point embedded systems
Author :
Bartl, Matous ; Cermak, Karel
Author_Institution :
Dept. of Appl. Electron. & Telecommun., Univ. of West Bohemia, Plzeň, Czech Republic
Abstract :
This paper describes the methodology for implementation of artificial neural networks with adaptable parameters (weights, connections, number of neurons) on fixed-point embedded systems. Components of neuron unit and interconnecting matrix are discussed. Particular example of implementation on PIC18F46K80 is given. Results are discussed in appropriate part.
Keywords :
embedded systems; fixed point arithmetic; microcontrollers; neural nets; PIC18F46K80; adaptable parameters; adaptable topology; artificial neural network implementation; connection parameter; fixed-point embedded systems; interconnecting matrix; neuron parameter; neuron unit components; weight parameter; Accuracy; Approximation methods; Artificial neural networks; Microcontrollers; Neurons; Topology; Vectors; Artificial neural network; PIC18F46K80; neuron unit;
Conference_Titel :
Telecommunications Forum Telfor (TELFOR), 2014 22nd
Conference_Location :
Belgrade
Print_ISBN :
978-1-4799-6190-0
DOI :
10.1109/TELFOR.2014.7034444
