Recognition of Handwritten Digits Using Computer Vision Preprocessor Based Combined Architecture of Self-Organizing Map And Backpropagation on MNIST Dataset

In this paper, we propose a neural network system of combined architecture, using a self-organizing map (SOM) along with the concept of backpropagation to recognize handwritten digits from the MNIST dataset. The handwritten digits of the MNIST dataset were processed through a computer vision pre-processor. The general problem with the backpropagation method, which is its big learning time for large datasets, is attempted to be removed when used with an unsupervised mode of classification such as SOM when the data being used for backpropagation is already made to go through the SOM algorithm.