A Novel Interpolation Perspective for Handwritten Digit Recognition using Neural Network

In this work, we present an innovative technique for manually written character recognition that is disconnected, using deep neural networks. Since of the accessibility of enormous knowledge calculation and numerous algorithmic advances that are emerging, it has become easier in this day and age to train deep neural systems. And seeks to classify the numerical digits so that digits can be translated into pixels. Today, the computing force measure required to prepare a neural system has increased owing to the proliferation of GPUs and other cloud-based administrations like Google and Amazon offer tools to prepare a cloud-based neural system. We also developed a system for the recognition of character dependent on manually written image division. This project uses libraries such as NumPy, pandas, sklearn, seaborn to accomplish this either by linear and non-linear algorithm, to know its precision on confusion matrix and accuracy. This idea spins with RBF(radial basis function) which consists of two parameters as C and Gamma and classifying the pixel digits. To train those models, research work includes Convolutional Neural Network (CNN), Dynamic Neural Network(DNN), Recurrent Neural Network(RNN), and TensorFlow algorithms using Keras , which can be accurately used for the classification of the digits.