PoolBasedSplitter<T>
Models & Types ยท AiDotNet.Preprocessing.DataPreparation.Splitting.ActiveLearning
Pool-based active learning splitter that maintains labeled and unlabeled pools.
Inheritance: DataSplitterBase<T> โ PoolBasedSplitter<T>
For Beginners
In pool-based active learning, we start with a small labeled dataset and a large pool of unlabeled data. The model queries samples from the unlabeled pool that it is most uncertain about, and an oracle (human expert) provides labels.
How It Works
How It Works:
- Start with a small initial labeled set (seed)
- The rest becomes the unlabeled pool
- Model queries uncertain samples from the pool
- Oracle labels them, moving them to the labeled set
When to Use:
- Limited labeling budget
- Expensive manual annotation
- Interactive machine learning
- Iterative model improvement
Constructors
| Constructor | Summary |
|---|---|
PoolBasedSplitter(double initialLabeledRatio, int minLabeledSamples, SelectionStrategy<T> strategy, bool shuffle, int randomSeed) |
Creates a new pool-based active learning splitter. |
Properties
| Property | Summary |
|---|---|
bool RequiresLabels { get; } |
|
string Description { get; } |
Methods
| Method | Summary |
|---|---|
DataSplitResult<T> Split(Matrix<T> X, Vector<T> y) |