Training Support Vector Machines (SVMs)

In this project, we’ll implement a support vector machine (SVM) from scratch on Iris, a popular toy dataset. The project will leverage the concepts of Lagrange duality and projected gradient descent (PGD).

The primary objective is to understand and apply the mathematical underpinnings of SVMs, particularly the optimization problem involving Lagrange multipliers. The challenge will involve formulating the primal problem, transforming it into its dual form using Lagrange duality, and solving the dual problem to find the optimal hyperplane.

This project will provide a deeper understanding of SVMs and their practical applications in machine learning.