Table of Contents

Class MaxPool3DLayer<T>

Namespace
AiDotNet.NeuralNetworks.Layers
Assembly
AiDotNet.dll

Represents a 3D max pooling layer for downsampling volumetric data.

public class MaxPool3DLayer<T> : LayerBase<T>, ILayer<T>, IJitCompilable<T>, IDiagnosticsProvider, IWeightLoadable<T>, IDisposable

Type Parameters

T

The numeric type used for calculations, typically float or double.

Inheritance
MaxPool3DLayer<T>
Implements
Inherited Members

Remarks

A 3D max pooling layer reduces the spatial dimensions (depth, height, width) of volumetric data while preserving the most prominent features. This helps reduce computational cost and provides translation invariance.

For Beginners: Max pooling works like summarizing a 3D region by keeping only the largest value.

Think of it like this:

  • You have a 3D grid of numbers
  • You divide it into small cubes (e.g., 2x2x2)
  • For each cube, you keep only the largest number
  • This makes your data smaller while keeping the important features

This is useful because:

  • It reduces the amount of computation needed
  • It helps the network focus on the most important features
  • It makes the network more robust to small position changes

Constructors

MaxPool3DLayer(int[], int, int)

Initializes a new instance of the MaxPool3DLayer<T> class.

public MaxPool3DLayer(int[] inputShape, int poolSize, int stride = 0)

Parameters

inputShape int[]

The shape of the input tensor [channels, depth, height, width].

poolSize int

The size of the pooling window (applied to all three dimensions).

stride int

The stride for moving the pooling window. Defaults to poolSize if 0.

Remarks

For Beginners: This creates a max pooling layer that shrinks 3D data.

If stride equals pool size, the output will have dimensions reduced by the pool size factor. For example, with pool size 2 and stride 2, a 32x32x32 input becomes 16x16x16.

Exceptions

ArgumentException

Thrown when inputShape has invalid length.

ArgumentOutOfRangeException

Thrown when poolSize or stride is non-positive.

Properties

PoolSize

Gets the size of the pooling window (same for depth, height, width).

public int PoolSize { get; }

Property Value

int

Remarks

The pool size determines the size of the region from which to extract the maximum value. Common values are 2 (halves each spatial dimension) or 3.

Stride

Gets the stride (step size) for moving the pooling window.

public int Stride { get; }

Property Value

int

Remarks

Stride controls how much the pooling window moves between positions. A stride equal to pool size produces non-overlapping regions. A stride of 1 with pool size > 1 produces overlapping regions.

SupportsGpuExecution

Gets a value indicating whether this layer supports GPU execution.

protected override bool SupportsGpuExecution { get; }

Property Value

bool

Remarks

MaxPool3D supports GPU execution via CUDA, OpenCL, and HIP backends.

SupportsJitCompilation

Gets a value indicating whether this layer supports JIT compilation.

public override bool SupportsJitCompilation { get; }

Property Value

bool

true if the input shape is configured.

SupportsTraining

Gets a value indicating whether this layer supports training (backpropagation).

public override bool SupportsTraining { get; }

Property Value

bool

Always true as MaxPool3D supports gradient flow through max indices.

Methods

Backward(Tensor<T>)

Performs the backward pass to route gradients through max pooling.

public override Tensor<T> Backward(Tensor<T> outputGradient)

Parameters

outputGradient Tensor<T>

The gradient of the loss with respect to this layer's output.

Returns

Tensor<T>

The gradient of the loss with respect to this layer's input.

Remarks

During backpropagation, gradients are routed only to the positions that had the maximum values in the forward pass. All other positions receive zero gradient.

Exceptions

InvalidOperationException

Thrown when Forward has not been called.

BackwardGpu(IGpuTensor<T>)

Performs GPU-resident backward pass of 3D max pooling.

public override IGpuTensor<T> BackwardGpu(IGpuTensor<T> outputGradient)

Parameters

outputGradient IGpuTensor<T>

The gradient of the output on GPU.

Returns

IGpuTensor<T>

The gradient with respect to input as a GPU-resident tensor.

Deserialize(BinaryReader)

Deserializes the layer from a binary stream.

public override void Deserialize(BinaryReader reader)

Parameters

reader BinaryReader

The binary reader to deserialize from.

ExportComputationGraph(List<ComputationNode<T>>)

Exports the layer as a computation graph for JIT compilation.

public override ComputationNode<T> ExportComputationGraph(List<ComputationNode<T>> inputNodes)

Parameters

inputNodes List<ComputationNode<T>>

List to populate with input nodes.

Returns

ComputationNode<T>

The output computation node.

Exceptions

ArgumentNullException

Thrown when inputNodes is null.

InvalidOperationException

Thrown when layer is not properly initialized.

Forward(Tensor<T>)

Performs the forward pass of 3D max pooling.

public override Tensor<T> Forward(Tensor<T> input)

Parameters

input Tensor<T>

The input tensor with shape [batch, channels, depth, height, width] or [channels, depth, height, width].

Returns

Tensor<T>

The pooled output tensor with reduced spatial dimensions.

Remarks

This method uses the vectorized Engine.MaxPool3DWithIndices operation for CPU/GPU acceleration. The max indices are cached for use in the backward pass.

Exceptions

ArgumentException

Thrown when input tensor has invalid rank.

ForwardGpu(params IGpuTensor<T>[])

Performs GPU-resident forward pass of 3D max pooling, keeping all data on GPU.

public override IGpuTensor<T> ForwardGpu(params IGpuTensor<T>[] inputs)

Parameters

inputs IGpuTensor<T>[]

The input tensors on GPU (uses first input).

Returns

IGpuTensor<T>

The pooled output as a GPU-resident tensor.

GetActivationTypes()

Gets the activation function types used by this layer.

public override IEnumerable<ActivationFunction> GetActivationTypes()

Returns

IEnumerable<ActivationFunction>

An empty enumerable as max pooling has no activation function.

GetParameters()

Gets all trainable parameters. Max pooling has none.

public override Vector<T> GetParameters()

Returns

Vector<T>

An empty vector.

ResetState()

Resets the cached state from forward/backward passes.

public override void ResetState()

Remarks

Call this method to free memory after inference is complete or when switching between different inputs.

Serialize(BinaryWriter)

Serializes the layer to a binary stream.

public override void Serialize(BinaryWriter writer)

Parameters

writer BinaryWriter

The binary writer to serialize to.

UpdateParameters(T)

Updates parameters. Max pooling has no trainable parameters.

public override void UpdateParameters(T learningRate)

Parameters

learningRate T

The learning rate (unused).

Edit this page