Automatic Recognition of Handwritten Digits Using Multi-Layer Sigmoid Neural Network

Said Kassim Katungunya, Xuewen Ding, Juma Joram Mashenene

Abstract: One of the challenge we face in human vision is recognizing handwritten digits, since digits writing by free hand may differ from one person to another, sometimes it can be difficult to identify exact type of digits due to their shape and style upon which the digit is written. Automatic recognition of handwritten digits using multi-layer sigmoid neural network provides a solution to this problem. This approach employs the use of sigmoid neuron with feedforward and backpropagation technique. This tool can ensure accuracy of more than 98 percent in recognizing various handwritten digits.

Keywords: logistic regression, sigmoid function, Feedforward, Backpropagation, regularization parameter

How to Cite?: Said Kassim Katungunya, Xuewen Ding, Juma Joram Mashenene, "Automatic Recognition of Handwritten Digits Using Multi-Layer Sigmoid Neural Network", Volume 5 Issue 3, March 2016, International Journal of Science and Research (IJSR), Pages: 951-955, https://www.ijsr.net/getabstract.php?paperid=NOV162037, DOI: https://dx.doi.org/10.21275/NOV162037