Major tasks performed in this assignment are-
-
Used TSNE (t-distributed stochastic neighbour embedding) to reduce the dimensions of the MNIST dataset and visualize it using scatter plot.
-
Implemented Linear Regression (Least Squares), Linear Regression with L1 regularization (LASSO) and L2 regularization (Ridge Regression) from scratch using gradient descent and used this implementation on the Abalone Dataset.
-
Used Logistic Regression (with and without regularization) from the sklearn library on the UCI Ionosphere Dataset and trained a model that predicted labels with an accuracy of 91%.
-
Used PCA (Principal Component Analysis) to reduce the dimensions of the MNIST dataset. Used OVO (One-vs-One) and OVR (One-vs-Rest) with Logistic Regression to predict the labels in MNIST dataset.
-
Evaluated the performance of the above models using various techniques like precision, recall, AUC-ROC (implemented from scratch), accuracy and f1-score.