Skip to main content

MaxPool

2D Max Pooling

Downsamples by taking the maximum value in each pooling window. Commonly used in CNNs to reduce spatial dimensions while preserving the strongest activations.

Parameters:

  • kernel_size: Size of the pooling window
  • stride: Stride of the pooling window (default: 1)
  • padding: Zero-padding added to input (default: 0)
  • dilation: Spacing between kernel elements (default: 1)

Shape Contract:

  • Input: [batch, channels, height, width]
  • Output: [batch, channels, out_height, out_width]

Notes:

  • Output size: floor((input + 2*padding - kernel_size) / stride) + 1
  • Common pattern: kernel_size=2, stride=2 (halves spatial dimensions)
  • Provides translation invariance
  • No learnable parameters
  • Preserves channel count

Signature​

neuron MaxPool(kernel_size, stride=Int(1), padding=Int(0), dilation=Int(1))

Ports​

Inputs:

  • default: [batch, channels, height, width]

Outputs:

  • default: [batch, channels, *, *]

Implementation​

Source { source: "core", path: "pooling/MaxPool" }