Graduate Students Seminar

Speaker 1: Jhilam Sur

Title

Breast cancer Prediction using Logistic Regression

Abstract

Logistic Regression is a useful regression model for predicting binary outcomes and calculating their corresponding probabilities. In the beginning there is a brief explanation of logistic regression model with its sigmoid curve. Breast cancer is a disease whose early detection is critical for patient care. I have investigated the Breast Cancer dataset by Wisconsin-Madison available on kaggle, to see which variables in the model are good for correct prediction of whether the tumor is malignant or benign. Discussed the summary statistics and compared between 2 models.

Speaker 2: Saeed Damadi

Title

Joint Group Sparsity Projection

Abstract

Solving an optimization problem with a sparsity constraint is difficult because it is nonconvex, requiring the solution of many subproblems. One approach is to consider a group sparsity constraint to identify the optimal group(s) with optimal components. To the best of our knowledge, there is no direct method in the literature for handling group constraints. We have developed an algorithm that solves an optimization problem with a group sparsity constraint. The algorithm is applied to synthetic linear and logistic regression problems and compared with the Lasso and Sparse Group Lasso algorithms.