Abstract: The human visual system is one of the wonders of the world. The difficulty of visual pattern recognition becomes apparent if you attempt to write a computer program to recognize digits. One of the approaches to computers behaving and computing like humans involve neural networks. The neural network automatically infers rules for recognizing handwritten digits. This approach reduces human intervention in many commercial places like banks to process cheques and by post offices to recognize addresses. The neural network attempts to determine if the input data matches a pattern that the neural network has memorized. The concepts of weights and biases are conventionally used to enhance the performance of the Neural networks over a large set of test images and classify the digit into a class label. The data is stored in a very simple file format designed for storing vectors and multidimensional matrices. Tensor Flow is used to train an elaborate model that achieves state-of-the-art performance along with Softmax Regression. The current program can recognize digits with an accuracy over 96 percent, without human intervention, classifying 9,979 of 10,000 images correctly. The performance is close to human-equivalent, and is arguably better, since quite a few of the MNIST images are difficult even for humans to recognize with confidence.
Keywords: Neural network; TensorFlow; Softwax Regression; Pattern recognition.