Dropout - Klyn API

Dropout

classin packageklyn.math.nn.layers

public class Dropout<T extends AbstractFloat = Float> extends Layer<T>

Dropout

Random element-wise dropout during training, identity during inference. Kept values are multiplied by 1/(1-rate). Each training call advances a layer-local seed; equal seeds reproduce the same sequence of masks. A layer is not safe for concurrent training. Saved masks belong to individual passes, so a second forward call does not overwrite the first backward context.

Example

model = Sequential<Float>([Dense<Float>(64), Dropout<Float>(0.25), Dense<Float>(10)])

Properties

Modifier and Type	Member	Description
public override readonly property	displayName `displayName as String`	Summary label.
public override readonly property	parameterCount `parameterCount as ULong`	No trainable parameters.
public readonly property	randomState `randomState as Long`	Current random state; restoring it reproduces the next mask.
public readonly property	rate `rate as Double`	Drop probability in [0, 1).

Constructors

Modifier and Type	Member	Description
public	Dropout `Dropout(rate as Double, seed as Long = 1l)`	Configures independent, reproducible dropout without touching global randomness.

Methods

Modifier and Type	Member	Description
public override	backwardTraining `backwardTraining(context as LayerPass<T>, upstream as NDArray<T>) as LayerGradients<T>`	Reuses exactly the keep flags of the corresponding forward pass.
public override	buildShape `buildShape(inputShape as IList<Long>) as IList<Long>`	Preserves any positive sample shape.
public override	differentiate `differentiate(inputs as NDArray<T>, upstream as NDArray<T>) as LayerGradients<T>`	Differentiates inference, where dropout is the identity.
public override	forward `forward(inputs as NDArray<T>) as NDArray<T>`	Returns inputs unchanged during inference, with no allocation or random draw.
public override	forwardTraining `forwardTraining(inputs as NDArray<T>) as LayerPass<T>`	Samples a fresh mask and retains it only in the returned context.
public override	parameter `parameter(index as Long) as NDArray<T>`	Rejects parameter access.