Overview > klyn.math.nn.layers > Dropout > differentiate
Version 0.2.1 Overview Index Tree Deprecated
klyn.math.nn.layers.Dropout.differentiate
method
public override differentiate(inputs as NDArray<T>, upstream as NDArray<T>) as LayerGradients<T>
Description

Differentiates inference, where dropout is the identity. Training callers must use backwardTraining with the saved forwardTraining context.

Parameters
ParameterDescription
inputsInference input.
upstreamSame-shaped derivative.
Returns

Upstream unchanged, without parameter derivatives.

Throws
  • ValueException for invalid input or gradient shapes.
Example

gradient = dropout.differentiate(batch, upstream)