Softmax - Klyn API

Softmax

classin packageklyn.math.nn.activations

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

Softmax

Stable row-wise normalization of finite logits into class probabilities. Use SparseCategoricalCrossentropy(fromLogits=false) after this activation. A loss configured with fromLogits=true instead requires a linear output.

Example

probabilities = Softmax<Float>().forward(model.predict(xTest))

Properties

Modifier and Type	Member	Description
public override readonly property	displayName `displayName as String`	Summary label.
public override readonly property	parameterCount `parameterCount as ULong`	Softmax has no trainable parameters.

Methods

Modifier and Type	Member	Description
public override	backwardTraining `backwardTraining(context as LayerPass<T>, upstream as NDArray<T>) as LayerGradients<T>`	Reuses saved probabilities rather than evaluating exponentials again.
public	build `build(inputSize as Long) as Long`	Checks and preserves a positive feature width.
public override	buildShape `buildShape(inputShape as IList<Long>) as IList<Long>`	Validates a vector of class scores.
public override	differentiate `differentiate(inputs as NDArray<T>, upstream as NDArray<T>) as LayerGradients<T>`	Computes a Jacobian-vector product without materializing a Jacobian.
public override	forward `forward(inputs as NDArray<T>) as NDArray<T>`	Computes exp(x - max(x)) / sum(exp(x - max(x))) along each row.
public override	parameter `parameter(index as Long) as NDArray<T>`	Rejects parameter access on this parameter-free activation.