This lesson introduces support vector machines (SVMs), their maximum-margin intuition, and the kinds of problems they can address.
A support vector machine is a supervised learning model associated with algorithms for classification and regression. In binary classification, training examples have features and labels, often encoded as +1 and −1. The model learns a decision boundary that can be used to assign a label to a new example.
All flashcards
What is a support vector machine?
A supervised learning model that constructs decision boundaries for tasks such as classification and regression.
What is the margin?
The separation between a decision boundary and its nearest training examples. Maximum-margin SVMs seek to make this separation large.
What is a support vector?
A training point on or near the margin that contributes to determining the learned boundary.
What is support vector clustering?
An unsupervised approach that applies support-vector ideas to unlabeled data to identify natural clusters.
Name three application areas for SVMs.
Examples include text categorization, image classification, satellite data analysis, handwriting recognition, and protein classification.
Quiz questions
What kind of learning model is an SVM?Answer: A supervised model commonly used for classification and adaptable to regression.
What does the dual formulation of a linear SVM optimize?Answer: Coefficients associated with the training examples.
How does sequential minimal optimization simplify SVM training?Answer: It solves small subproblems involving two variables at a time.
In a one-versus-one multiclass SVM strategy, what classifiers are trained?Answer: One classifier for each pair of classes.
Which is a limitation of standard SVMs?Answer: They do not directly provide calibrated class-membership probabilities.