Table of Contents#### Download Safari Books Online apps: Apple iOS | Android | BlackBerry

### 6.7. Summary

Entire Site

Free Trial

Safari Books Online is a digital library providing on-demand subscription access to thousands of learning resources.

Support vector machines are a type of classifier. They’re called machines because they generate a binary decision; they’re decision machines. Support vectors have good generalization error: they do a good job of learning and generalizing on what they’ve learned. These benefits have made support vector machines popular, and they’re considered by some to be the best stock algorithm in unsupervised learning.

Support vector machines try to maximize margin by solving a quadratic optimization problem. In the past, complex, slow quadratic solvers were used to train support vector machines. John Platt introduced the SMO algorithm, which allowed fast training of SVMs by optimizing only two alphas at one time. We discussed the SMO optimization procedure first in a simplified version. We sped up the SMO algorithm a lot by using the full Platt version over the simplified version. There are many further improvements that you could make to speed it up even further. A commonly cited reference for further speed-up is the paper titled “Improvements to Platt’s SMO Algorithm for SVM Classifier Design.”^{[6]}

^{[6]}S. S. Keerthi, S. K. Shevade,. Bhattacharyya, and K. R. K. Murthy, “Improvements to Platt’s SMO Algorithm for SVM Classifier Design,” Neural Computation 13, no. 3, (2001), 637–49.