OnlineKMeans<T>

Models & Types · AiDotNet.Clustering.Streaming

Online K-Means clustering for streaming data.

Inheritance: ClusteringBase<T> → OnlineKMeans<T>

For Beginners

Online K-Means learns continuously.

Think of it like learning a new skill:

  • Each example you see teaches you a little bit
  • You don’t need to remember every example
  • Your understanding improves over time

The learning rate is like “how much you learn from each example”:

  • High: Quick learner, but might forget old lessons
  • Low: Slow learner, but very stable knowledge

Use cases:

  • Real-time sensor data
  • Social media streams
  • Financial market data
  • Any data that arrives continuously

How It Works

Online K-Means processes data points one at a time, updating cluster centers incrementally. This is suitable for streaming data or when memory is limited.

Algorithm:

  1. Initialize K cluster centers
  2. For each incoming point:

a. Find the nearest cluster center b. Move that center slightly toward the point c. Update cluster assignment

  1. Optionally decay learning rate over time

Example

using AiDotNet;
using AiDotNet.Data.Loaders;
using AiDotNet.Clustering.Streaming;
using AiDotNet.Tensors.LinearAlgebra;

var data = new Matrix<double>(6, 2);
double[][] rows = { new[] { 1.0, 1.0 }, new[] { 1.2, 0.9 }, new[] { 1.1, 1.1 },
                    new[] { 8.0, 8.0 }, new[] { 8.2, 7.9 }, new[] { 7.9, 8.1 } };
for (int i = 0; i < 6; i++) { data[i, 0] = rows[i][0]; data[i, 1] = rows[i][1]; }

var result = await new AiModelBuilder<double, Matrix<double>, Vector<double>>()
    .ConfigureModel(new OnlineKMeans<double>())
    .ConfigureDataLoader(DataLoaders.FromMatrix(data))
    .BuildAsync();

var labels = result.Predict(data);
Console.WriteLine($"OnlineKMeans: clustered {labels.Length} points.");

Constructors

Constructor Summary
OnlineKMeans(OnlineKMeansOptions<T> options) Initializes a new Online K-Means instance.

Properties

Property Summary
double CurrentLearningRate { get; set; } Gets the current learning rate.
long TotalPointsSeen { get; } Gets the total number of points seen during training.

Methods

Method Summary
IFullModel<T, Matrix<T>, Vector<T>> WithParameters(Vector<T> parameters)
ModelOptions GetOptions()
Vector<T> FitPredict(Matrix<T> x)
Vector<T> Predict(Matrix<T> x)
int PartialFit(Vector<T> point) Processes a single data point (true online/streaming mode).
void InitializeWithRandom(int numFeatures, T[] minValues, T[] maxValues) Initializes centers randomly for streaming mode.
void Train(Matrix<T> x, Vector<T> y)