MiniBatchKMeansOptions<T>

Options & Configuration · AiDotNet.Clustering.Options

Configuration options specific to MiniBatch K-Means clustering.

Inheritance: ModelOptions → ClusteringOptions<T> → MiniBatchKMeansOptions<T>

For Beginners

MiniBatch K-Means is like regular K-Means but faster.

Instead of using all data points in every step, it uses a random sample (mini-batch). This makes it:

  • Much faster for large datasets (millions of points)
  • Uses less memory
  • Produces slightly less optimal clusters (but usually very close)

Use MiniBatch K-Means when:

  • Your dataset has more than ~10,000 points
  • Speed is more important than perfect clustering
  • You’re doing online/streaming clustering

How It Works

MiniBatch K-Means is a variant of KMeans that uses mini-batches (random samples) to reduce computation time for large datasets while producing similar results.

Constructors

Constructor Summary
MiniBatchKMeansOptions()

Properties

Property Summary
KMeansInitMethod InitMethod { get; set; } Gets or sets the initialization method for cluster centers.
Matrix<T> InitialCenters { get; set; } Gets or sets custom initial cluster centers.
bool ReassignEmptyClusters { get; set; } Gets or sets whether to reassign empty clusters.
double ReassignmentRatio { get; set; } Gets or sets the fraction of centers that must be reassigned in each iteration.
int BatchSize { get; set; } Gets or sets the size of the mini-batches.
int MaxNoImprovement { get; set; } Gets or sets the number of iterations with no improvement to wait before stopping.
int NumClusters { get; set; } Gets or sets the number of clusters to find.
int? InitSize { get; set; } Gets or sets the number of random samples used to initialize centers.