Spectral Clustering

What is Spectral Clustering?

Spectral clustering is an exploratory data analysis technique that reduces complex multidimensional datasets into clusters of similar data in fewer dimensions. The goal is to cluster the full spectrum of unorganized data points (the eigenvalues) into several groups based upon their similarity. This groups similar data, regardless of features, around common points.

How is Spectral Clustering Used in Machine Learning?

Due to its simple implantation and relatively low performance requirements compared to other clustering techniques, spectral clustering is one of the most popular forms of multivariate statistical analysis.

There are three main steps to perform spectral clustering:

Create a similarity matrix between each pair of points or N objects in the dataset to use as an input.

Find the k eigenvectors of the Laplacian matrix, which defines the feature vector for each object.

Run a k-means function on the new features to separate objects into k-classes.