/pegasos

Implements modified version of the Pegasos (Primal Estimated Sub-Gradient Solver for SVM) algorithm as well as Perceptron and Average Perceptron for comparison

Primary LanguagePython

pegasos

An implementation of the Pegasos algorithm with hinge loss: "a simple and effective stochastic sub-gradient descent algorithm for solving the optimization problem cast by Support Vector Machines (SVM)" (Shalev-Shwartz et al.).

About

Read the original paper on the Pegasos (Primal Estimated Sub-Gradient Solver for SVM) here.

Implementation

The algorithm was implemented in Python, and the Perceptron and Average Perceptron algorithms were also implemented as a comparison. Some functions for bag-of-words functionality are included as wel.