Study of Lasso and Ridge Regression using ADMM

Convex optimization has several applications, in-cluding communication networks, data analytics, economics, and statistics. This technique emerged as the most efficient method of optimization. It can be performed using a variety of algorithms like gradient descent, penalties, augmented Lagrangians, Bundles, etc. In this paper, an optimization method for convex problems called ADMM has been discussed. The optimization of Lasso and Ridge regression using ADMM was part of our study, as well as a comparison of the ADMM implementation with the original model from Scikit-Learn's machine learning library and the gradient descent optimization approach for the same problems on the Diabetes dataset. We then compared these models using certain key evaluation metrics. Furthermore, we examined the performance of the three models and discovered that ADMM is an efficient algorithm as it converges to the optimum faster.