Skip to main content

GlobalAvgPool

Global Average Pooling

Reduces each channel to a single value by averaging all spatial positions. Commonly used as the final pooling before classification head.

Shape Contract:

  • Input: [batch, channels, height, width] (any spatial size)
  • Output: [batch, channels, 1, 1] single value per channel

Notes:

  • Equivalent to AdaptiveAvgPool(1)
  • Reduces spatial dimensions to 1x1
  • No learnable parameters
  • Reduces overfitting compared to large fully-connected layers
  • Works with any input resolution
  • Standard in ResNet, EfficientNet, and modern CNNs

Signature​

neuron GlobalAvgPool()

Ports​

Inputs:

  • default: [batch, channels, *, *]

Outputs:

  • default: [batch, channels, 1, 1]

Implementation​

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