00000ctmaa22000003a 4500 UP-1685675941123923467 Buklod 20151103143253.0 aa rb |||1 u| ta 151103s xx d r |||| u| (iLib)UPTAC-00017804977 UPVTC eng LG 993.5 2008 C66 C34 Cagadoc, Mark Victor E. author. NetSCORe neural network using Scaled Conjugate Gradient in optical character recognition Mark Victor E. Cagadoc [and] Marie Claire B. Royeras; John Paul T. Yusiong, adviser. 2008. [4], 65 leaves illustrations. Undergraduate thesis (B.S. Computer Science) -- University of the Philippines, Tacloban. Character Recognition is a field of research in pattern recognition and Artificial Intelligence that is essential to document management, form processing and any other commercial applications. Generally the most common approach to optical character recognition problem is the application of neural network. Neural Networks are used for pattern recognition or data classification through the learning process called training. There are already successful applications of certain training algorithms such as Back Propagation to character recognition. However, they also have shortcomings for example slow training process. One possible remedy to this is to use a better training algorithm. In this study, the Scaled Conjugate Gradient algorithm (SCG) is used as an alternative for training artificial neural networks. This algorithm belongs to the class of Conjugate Gradient Methods, which show super linear convergence on most problems. Experiments have shown that the Scaled Conjugate Gradient Algorithm is an effective learning algorithm especially when applied to Character Recognition problem. It uses a step size scaling mechanism which avoids time consuming line-- search per learning iteration that makes the algorithm faster than other training algorithms. Character recognition. Scaled Conjugate Gradient Algorithm. Neural network. Training algorithms. Royeras, Marie Claire B. author. Yusiong, John Paul T. adviser. FI UP UPTAC UPTAC LG 993.5 2008 C66 C34 Thesis