liveBook · Manning

Linear kernel

As explained in Timeless Algorithms, The Foundational Ideas

A linear kernel computes similarity using the ordinary dot product of observations in the supplied feature space.

In a support vector machine (SVM), the linear kernel compares observations with the ordinary dot product, K(xᵢ, x) = xᵢ · x. It uses the supplied feature space without mapping observations into a richer one. section 8.6.2

K(xi,x)=xi∙x

The kernel value is the dot product between observations xᵢ and x. section 8.6.2

The SVM’s resulting boundary is a hyperplane: a line in two dimensions, a plane in three, and a flat boundary in higher dimensions. The kernel does not determine the fitted boundary alone; training data, support vectors, and the penalty parameter also matter. Linearity refers to the supplied feature space: adding temperature squared as a feature can make a boundary linear in that expanded space while appearing nonlinear in temperature alone. section 8.6.2

How Timeless Algorithms, The Foundational Ideas covers it

How deep it goes
Defines the dot-product kernel and its boundary geometry, but does not map observations into a richer feature space.
Examples
Uses temperature and temperature squared to show how linearity depends on the supplied feature space.
What you can do after
Evaluate whether a flat boundary can separate the classes adequately or whether the supplied features are sufficiently rich.
How it connects
Links the SVM to dot products, feature spaces, support vectors, and hyperplanes.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Timeless Algorithms, The Foundational Ideas. The section links point to the parts of the book it draws on.