Libraries for Assignments: Numpy, MatplotLib, Pandas, Seaborn

Assignment-1: Solved problems related to error probabilities and decision boundaries in classification u, computing an unbiased covariance matrices, and deriving likelihood ratio tests.

Assignment-2: Implemented PCA and QDA from scratch on the MNIST dataset

Assignment-3: Performed PCA on selected MNIST classes, built a decision tree with 3 terminal nodes using Gini index, classified test samples, and improved results using bagging. All from scratch.

Assignment-4: Implemented AdaBoost.M1 on MNIST classes using PCA-reduced data, trained 300 decision stumps, tracked validation accuracy, and reported test accuracy; then, applied gradient boosting with absolute loss for regression, tracked MSE on validation set, and reporte test MSE. All from scratch

Project: Classified Dialogues to Characters in Seinfeld and The Office using a combination of NLP techniques (TF-IDF, NLTK Tokenisation), and compared the performances of MNB,MLP,RF,LR,VC and ANN based classifiers on the same, with a max classification accuracy of 74% on the latter, using dialogues from The Office.