The Recognition of Handwritten Digits Using Neural Network Technology

In this paper, we compared the quality of processing handwritten digits by a fully connected feedforward neural network and a convolutional neural network. Examples of training and use of both a fully connected neural network and a convolutional neural network are given. The ways of image segmentation and finding the boundaries of handwritten digits for their further analysis are considered. Algorithms affecting the accuracy of learning, the speed of learning are considered. The general principles of the structure and principles of operation of a feed-forward neural network and a convolutional neural network are presented.