Gist Machine Learning Course Coding Assignment

SVM (Support Vector Machine) Implementation

• Course: GIST Machine Learning (EC4213)
• Project Type: SVM Implementation Individual Coding Assignment

Description

An educational project focusing on the implementation of Support Vector Machines (SVM) from scratch. The project covers key concepts such as hard margin SVM, soft margin SVM (primal and dual formulations), and the use of kernel tricks to handle non-linear data.

• SVM_hard.py : Implementation of SVM with hard margin.
• SVM_soft.py : Implementation of SVM with soft margin.
• SVM_kernel.py : Implementation of kernels which will be used to soft margin.
• utils.py : A bunch of utility functions!
• test.py : A testing code! We run this code to evaluate implementation.

Overview

Hard margin SVM

• We implemented the process of finding the optimal decision boundary using hinge loss and coordinate gradient descent.

Soft margin SVM

• We can find a decision boundary of two classes by solving dual problem. Slack variables allow misclassification.

Kernel Tricks

• We implemented various kernel filters to SVM to compare their performance.