SequenceLastLayer<T>

Layers ยท AiDotNet.NeuralNetworks.Layers

A layer that extracts the last timestep from a sequence.

Inheritance: LayerBase<T> โ†’ SequenceLastLayer<T>

Implements: IShapeContract

For Beginners

When processing sequences (like sentences or time series), recurrent layers output a value for each timestep. For tasks like classification, we often only need the final output (after seeing the whole sequence). This layer extracts just that last output.

How It Works

This layer is used after recurrent layers (RNN, LSTM, GRU) when the task requires a single output from the entire sequence, such as sequence classification.

Constructors

Constructor Summary
SequenceLastLayer(int featureSize) Initializes a new SequenceLastLayer.

Properties

Property Summary
bool SupportsGpuTraining { get; }
bool SupportsTraining { get; } Gets a value indicating whether this layer supports training.

Methods

Method Summary
IReadOnlyList<OutputAxisContract> OutputAxesFor(int inputRank)
void ResetState() Reset state is a no-op since this layer maintains no state between forward passes.