Implementation of a Support Vector Machine using the Sequential Minimal Optimization (SMO) algorithm for training.

Implementation of Platt's SMO algorithm (Platt, John (1998). "Sequential Minimal Optimization: A Fast Algorithm for Training Support Vector Machines") with improvements by Keerthi (S.S. Keerthi. S.K. Shevade. C. Bhattacharyya &. K.R.K. Murthy (2001). "Improvements to Piatt's SMO algorithm for SVM classifier design").

• NumPy

Following parameters are default

from SVM import SVM

model = SVM(C=1.0, kernel='rbf', degree=3, gamma='auto', coef0=0.0, tol=1e-3, max_iter=-1)

Parameters have the same meaning as parameters in SVC of scikit-learn package (link). But parameter kernel must be one of ‘linear’, ‘poly’, ‘rbf’ or a callable.

This implementation only for two classes: -1 and 1.

Since Keerthi's article proposes two training methods, there are two modifications to this implementation.

First method:

model.fit_modification1(X_train, y_train)

and second method:

model.fit_modification2(X_train, y_train)

Method for classification of samples.

y_hat = model.predict(X_test)

Method for getting a value of w * x - b.

decision = model.decision_function(X)

Method for getting the accuracy on the given data and labels.

score = model.score(X, y)