In machine learning, a margin often refers to the distance between the two hyperplanes that separate linearly-separable classes of data points.
![In this image from Wikipedia, the dotted lines represent the two hyperplanes dividing the white and black data points. The region between the lines is the margin.](/images/margin.png)
The term is most commonly used when discussing support vector machines, but often appears in other literature discussing boundaries between points in a vector space.