Comparison of Support Vector Machines and Artificial Neural Networks
for Handwritten Digit Recognition
The field of Handwritten Digit Recognition and Optical Character Recognition have always found a place in researches dealing with machine learning and classification. Pre-processes, feature extraction and learning/classification are some proposed methods for handwritten digit recognition, and some other various approaches are also presented in certain databases.
Within these approaches Artificial Neural Network takes place either extensively or as a hybrid method whereas Support Vector Machine method that has been put forth recently, comes across as a successful method in handwritten digit recognition operations.
Nowadays, the use of Support Vector Machines for solving classification problems and feature extraction operations has become a widespread method. This method can classify non-linear data with a high percentage of accuracy. Although Support Vector Machines have a slower learning and classification process when compared with Artificial Neural Networks, they show up as an alternative method in areas where Artificial Neural Network is being used. Today mostly hybrid applications that combine Artificial Neural Networks and Hidden Markov Models are widely being used in handwriting and handwritten digit recognition. The aim of this thesis is to compare Support Vector Machines with Artificial Neural Network in handwritten digit recognition, and to present that it is possible to use Support Vector Machines as an alternative method in handwritten digit recognition processes.
DEVELİ Ahmet
Tez Adı : Zaman Dizisi Verilerinde Ani Değişimlerin Tahmini için Veri Madenciliği Yöntemleri ile Öznitelik Seçimi
Danışman : Doç. Dr. Olcay KURŞUN
Anabilim Dalı : Bilgisayar Mühendisliği
Programı : -
Mezuniyet Yılı : 2013
Tez Savunma Jürisi : Doç. Dr. Olcay KURŞUN
Prof. Dr. Ahmet SERTBAŞ
Yrd. Doç. Dr. Tolga ENSARİ
Yrd. Doç. Dr. Fatih KELEŞ
Yrd. Doç. Dr. Niyazi KILIÇ
Dostları ilə paylaş: |