Recognition of Handwritten Digit using Convolutional Neural Network in Python with Tensor flow and Comparison of Performance for Various Hidden Layers
Abstract : In recent times, with the increase of Artificial Neural Network (ANN), deep learning has brought a dramatic twist in the field of machine learning by making it more artificially intelligent. Deep learning is remarkably used in vast ranges of fields because of its diverse range of applications such as surveillance, health, medicine, sports, robotics, drones, etc. In deep learning, Convolutional Neural Network (CNN) is at the center of spectacular advances that mixes Artificial Neural Network (ANN) and up to date deep learning strategies. It has been used broadly in pattern recognition, sentence classification, speech recognition, face recognition, text categorization, document analysis, scene, and handwritten digit recognition. The goal of this paper is to observe the variation of accuracies of CNN to classify handwritten digits using various numbers of hidden layers and epochs and to make the comparison between the accuracies. For this performance evaluation of CNN, we performed our experiment using Modified National Institute of Standards and Technology (MNIST) dataset. Further, the network is trained using stochastic gradient descent and the backpropagation algorithm.
? Artificial neural network (ANN) consists of one input layer, one output layer and some layers which exist in between input layer and output layer, these middle layers are hidden layers.
? These days, an ever-increasing number of individuals use pictures to transmit data. It is additionally main stream to separate critical data from pictures. Image Recognition is an imperative research area for its generally used applications.
? In general, the field of pattern recognition, one of the difficult undertakings is the precise computerized recognition of human handwriting.
? The method is also analyzed and compared with existing techniques for road scene and indoor understanding.
? In Stochastic Gradient Descent (SDG) a small number of iteration will find effective solutions for the optimization problems.
? Moreover, in SDG, a small number of iteration will lead to a suitable solution.
? It has a powerful impact on many fields. Even, in nano-technologies like manufacturing semiconductors, CNN is used for fault detection and classification.
? Handwritten digit recognition has become an issue of interest among researchers.
? Researchers are working on this issue to reduce the error rate as much as possible in handwriting recognition.
• The proposed neural system was trained and tested on a dataset achieved from MNIST.
• In work was proposed a simple technique for vastly expanding the training set on base of elastic distortions.
• The purpose of the proposed work is to achieve comparable accuracy using a pure CNN architecture through extensive investigation of the learning parameters in CNN architecture for MNIST digit recognition.
• The proposed method used max-pooling indices of a feature map while decoding and observed good performance.
• It can be observed that our CNN model outperforms the various similar CNN models proposed by various researchers using the same MNIST benchmark dataset.
? The signal propagates back to the system, again and again, to update the shared weights and biases in all the receptive fields to minimize the value of cost function which increases the network’s performance.
? It is pretty challenging to get a good performance as more parameters are needed for the large-scale neural network.
? It randomly switches off some neurons during training to improve the performance of the network by making it more robust.
? Therefore, to enhance the performance of the network, a stochastic version of the algorithm is used.
? Comparing with their above performances based on MNIST dataset for the purpose of digit recognition we have achieved better performance for the CNN.
|