PARALEL BLOK FAKTORISASI QR DALAM SISTEM MEMORI TERSEBAR MULTIKOMPUTER BERBASIS MPI-LINUX
Mochamad Hariadi, Mauridhy Hery Purnomo · Jurnal Informatika (Petra Christian University) · 2007
Artificial neural networks (ANN) implementation in image classification problem requires a lot of training time which caused by enormous data size.This kind of data is known as raw data.An image data is extracted directly without any preprocessing.Many feature extraction techniques are offered to reduce the time consumed in training image data.Boolean function algorithm (BFA) in ANN (first and second approach) produces identical output values that can be used as feature of an image data.This feature can then be used as input in Back Propagation ANN that is known as the best ANN method in problems classification.The experiment shows that BFA can be used well as one of feature extraction methods.The second approach BFA in Back Propagation ANN shows better performance in recognition and is lesser time consuming in training than the first approach.