AdjustedMutualInformation<T>

Models & Types · AiDotNet.Clustering.Evaluation

Adjusted Mutual Information (AMI) for comparing clustering results.

Implements: IExternalClusterMetric<T>

For Beginners

AMI is like NMI but corrected for luck.

Regular NMI can be high just by chance if you have many clusters. AMI asks “How much better than random?”

  • AMI = 0: No better than random
  • AMI = 1: Perfect agreement
  • AMI can be negative if worse than random

How It Works

AMI adjusts NMI for chance agreement. Random clusterings will have AMI close to 0, while perfect agreement gives AMI = 1.

Constructors

Constructor Summary
AdjustedMutualInformation() Initializes a new AdjustedMutualInformation instance.

Methods

Method Summary
double Compute(Vector<T> trueLabels, Vector<T> predictedLabels)