TransposeLayer<T>

Layers · AiDotNet.NeuralNetworks.Layers

Reorders the axes of the input tensor according to a fixed permutation. Zero-parameter utility layer, primarily used to expose a different axis as the “last” dimension so a DenseLayer can operate on it (enables MLP-Mixer-style cross-axis MLPs without bespoke kernels).

Inheritance: LayerBase<T> → TransposeLayer<T>

Implements: IShapeContract

How It Works

The permutation passed to the constructor uses logical axis indices (excluding the batch axis). For rank-N inputs with a batch axis at position 0, this layer keeps the batch axis at index 0 and permutes the remaining N-1 axes per permutation.

Common pattern (MLP-Mixer temporal mixer):

Constructors

Constructor Summary
TransposeLayer(int[] permutation) Initializes a new TransposeLayer.

Properties

Property Summary
bool SupportsTraining { get; }

Methods

Method Summary
IReadOnlyList<OutputAxisContract> OutputAxesFor(int inputRank)
void ResetState()