Learning high-level features by deep Boltzmann machines for handwriting digits recogintion
Shuo Zhang, Wuyi Zhang, Kary Kang · 2014
Handwriting recognition is the ability of a computer to understand handwritten inputs from users. Generally it includes preprocessing, feature extraction, and classifier training. In this paper, we will develop a handwriting digit recognition system by using Deep Boltzmann Machine (DBM) together with the Support Vector Machine (SVM). DBM is a deep learning technique to learn high level features from the training data, while SVM is a method to train non-linear classifiers from the learn features. Such a framework is a promising way to build up a powerful digit recognition system. Our experimental result shows that our system can achieve desired performance.