Assignment Task

A. It is easy for the human to perform a task accurately by practicing it repeatedly and memorizing it for the next time. Human brain can process and analyse images easily. Also, recognize the different elements present in the images.

In this assignment, the goal is to correctly classify digits from a dataset of tens of thousands of handwritten images and build a classifier to predict the label of a given digit.

MNIST (“Modified National Institute of Standards and Technology”) is the de facto “hello world” dataset of computer vision. Since its release in 1999, this classic dataset of handwritten images has served as the basis for benchmarking classification algorithms. As new machine learning techniques emerge, MNIST remains a reliable resource for researchers and learners alike.

The dataset consists of images of handwritten numeric digits between 0-9. Each image is of 28 x 28 pixels, i.e. 28 pixels along both length and breadth of the image. Each pixel is an attribute with a numeric value (representing the intensity of the pixel), and thus, there are 784 attributes in the dataset. You can download the dataset from Kaggle here.

Design an appropriate deep L-layer neural network architecture to correctly classify digits from the MNIST dataset by choosing appropriate number of hidden layers and appropriate number of neurons in each layer.

• Write a Python program that implements the above designed architecture to identify digits from the MNIST dataset by implementing forward propagation and backward propagation. Tabulate the results obtained in terms of accuracy and error for training and test data
• Perform hyperparameter tuning to increase the prediction accuracy of the neural network and tabulate the results obtained for different settings.

B. For the MNIST data set described in Part A, in this assignment build a CNN model to predict the label of a given digit

• Write a Python program that implements CNN model to identify digits from the MNIST dataset. Tabulate the results obtained in terms of accuracy and error for training and test data.
• Perform hyperparameter tuning to increase the prediction accuracy of the neural network and tabulate the results obtained for different settings. Compare the results obtained from neural network and CNN models