Introduction
A support vector machine (SVM) is a powerful and versatile model for classification, regression and even outlier detection. It shines on small-to-medium datasets (hundreds to tens of thousands of examples) and on complex but well-defined problems — text classification, image features, bioinformatics.
The core idea is easy to picture. Imagine two groups of houses on either side of a valley, and you must build a road between them. Many roads would separate the two groups — but the safest one runs right down the middle, as far as possible from the houses on both sides. If a new house is built near the edge later, the wide road still leaves room for error.
An SVM finds that widest road. This topic explains the idea; the next ones extend it to messy data, curved boundaries and regression.