o LibrarieslikeSckitlearn,Tensorflow Keras and Pytorch should not be used for this assignment. o Usage of Numpy,Pandas and Matplotlib libraries is allowed.
Learning Task 1: Build a classifier (Perceptron Model - PM1) using the perceptron algorithm. Figure out whether the data set is linearly separable by building the model. By changing the order of the training examples, build another classifier (PM2) and outline the differences between the models – PM1 and PM2. Learning Task 2: Build a classifier (Perceptron Model - PM3) using the perceptron algorithm on the normalized data and figure out the difference between the two classifiers (PM1 and PM3). Learning Task 3: Change the order of features in the dataset randomly. Equivalently speaking, for an example of feature tuple (f1, f2, f3, f4, . . , f32), consider a random permutation (f3, f1, f4, f2, f6, ....., f32) and build a classifier (Perceptron Model – PM4). Would there be any change in the model, PM4, as compared to PM1. If so, outline the differences in the models and their respective performances.
Learning Task 1: Build Fisher’s linear discriminant model (FLDM1) on the training data and thus reduce 32dimensional problem to univariate dimensional problem. Find out the decision boundary in the univariate dimension using generative approach. You may assume gaussian distribution for both positive and negative classes in the univariate dimension. Learning Task 2: Change the order of features in the dataset randomly. Equivalently speaking, for an example of feature tuple (f1, f2, f3, f4, . . , f32), consider a random permutation (f3, f1, f4, f2, f6, ....., f32) and build the Fisher’s linear discriminant model (FLDM2) on the same training data as in the learning task 1. Find out the decision boundary in the univariate dimension using generative approach and you may assume gaussian distribution for both positive and negative classes in the univariate dimension. Outline the difference between the models – FLDM1 and FLDM2 - and their respective performances.
Learning Task 1: Build a classification model (LR1) using Logistic Regression. What happens to testing accuracy when you vary the decision probability threshold from 0.5 to 0.3, 0.4, 0.6 and 0.7. Learning Task 2: You should apply Feature Engineering Task 1 and Feature Engineering Task 2 and then build a classification model (LR2) using Logistic Regression. What happens to testing accuracy when you vary the decision probability threshold from 0.5 to 0.3, 0.4, 0.6 and 0.7. Learning Task 2: You should apply Feature Engineering Task 1 and Feature Engineering Task 2 and then build a classification model (LR2) using Logistic Regression. What happens to testing accuracy when you vary the decision probability threshold from 0.5 to 0.3, 0.4, 0.6 and 0.7