ConvergenceCriterion<T>

Models & Types · AiDotNet.ActiveLearning.StoppingCriteria

Stopping criterion based on learning curve convergence.

Implements: IStoppingCriterion<T>

For Beginners

This criterion analyzes the learning curve (how performance changes with more data) and stops when adding more data shows diminishing returns.

How It Works

How It Works:

Key Insight: Learning curves often follow a power law: performance = a - b*n^(-c) where n is the number of samples. As n grows, improvement decreases.

When to Use:

Reference: Figueroa et al. “Predicting sample size required for classification performance” (BMC 2012)

Constructors

Constructor Summary
ConvergenceCriterion() Initializes a new Convergence criterion with default parameters.
ConvergenceCriterion(int minHistory, double minImprovement, int lookAhead) Initializes a new Convergence criterion with specified parameters.

Properties

Property Summary
T PredictedImprovement { get; } Gets the predicted improvement from additional samples.
string Description { get; }
string Name { get; }

Methods

Method Summary
T GetProgress(ActiveLearningContext<T> context)
bool ShouldStop(ActiveLearningContext<T> context)
void Reset()