Support vector machine (SVM)
Supervised classifier that finds a maximum-margin boundary between classes, often extended to nonlinear problems via the kernel trick.
A support vector machine (SVM) is a supervised learning method that separates classes with a decision boundary chosen to maximise the margin, the distance to the nearest training points on each side. Only those borderline examples, called support vectors, determine the final model, which yields a sparse solution that often generalises well on small datasets. Corinna Cortes and Vladimir Vapnik published the widely cited formulation in 1995, building on earlier optimal-margin work at AT&T Bell Laboratories.
When classes are not linearly separable in the raw input space, SVMs use the kernel trick to compute inner products in a higher-dimensional feature space without building the full mapping explicitly. During the second AI winter, SVMs were a dominant strand of classical machine learning alongside ensembles and neural networks that had fallen out of favour. See the article and timeline entry.