GSA–Based Training of Convolutional Neural Networks for OCR Applications
Lucian‐Ovidiu Fedorovici, Radu‐Emil Precup, Radu‐Codruţ David, Florin Dr ̆agan · Atlantis computational intelligence systems · 2012
The chapter proposes a new training algorithm to train Convolutional Neural Networks (CNNs) in Optical Character Recognition (OCR) applications. This algorithm is characterized by the use of Gravitational Search Algorithms (GSAs) in combination with the back-propagation algorithm (BP), and a new family of CNNs is thus suggested.