MNIST Classification using various machine learning algorithms Execute the Main file 'Main.m' to execute all Questions 1-4 and plot their results

Input the MNIST data in the following form for PCA, LDA, SVM: t10k-images.idx3-ubyte train-images.idx3-ubyte t10k-labels.idx1-ubyte train-labels.idx1-ubyte

Input for CNN: 'data/imdb_mnist.mat' Previously trained data: 'Nets/lenet.mat'

The following is done in the main file 'Main.m':

a)Load data by calling mnist_parse.m b)Implement PCA, call PCA.m c)Implement LDA, call LDA.m d)Implement SVM, call svm.m e)Implement CNN, call cnn_train.m, inference.m

Function files:

a)mnist_parse.m = Load MNIST data and convert to matrices of training, testing data and labels b)PCA.m : Implements PCA, calls KNN_classifier.m for nearest neighbor classification c)LDA.m : Implements LDA d)svm.m : Implements SVM using LibSVM (will not work without libsvm) e)cnn_train.m : Creates the architecture for CNN f)inference.m : Checks the testing data using CNN trained model g)KNN_classifier.m : Implements nearest neighbour algorithm

NOTE: 1) If exectution of 1 function is required, please comment all subsequent questions to decrease running time. 2) For execution of each function, go inside the respective folder 3) Current value of K for nearest neighbor implementation is kept 1 for fast implementation. Actual value used is K = 5 (for the report).