Max pooling

As explained in Sutskever's List

A downsampling operation that retains the largest activation within each local spatial window.

Max pooling downsamples a feature map by keeping the strongest activation in each local spatial window. In AlexNet, this reduces computation and makes representations more robust to small positional changes and input distortions. section 2.5.1

AlexNet

AlexNet uses 3 × 3 max-pooling windows with stride 2 after its initial convolutional downsampling, reducing spatial dimensions further. section 2.5.1

Figure 2.3 The AlexNet architecture (cropped in the original paper). The numbers indicate spatial dimensions and channel depth at each stage. The input image (224 × 224 × 3 RGB) passes through a series of convolutional and pooling layers that progressively reduce spatial resolution (224→55→27→13) while increasing channel depth. Early downsampling is driven by the first convolutional layer (11 × 11 filters with a stride of 4), followed by max pooling layers (3 × 3 windows with a stride of 2) that further reduce spatial dimensions. The stride determines how far a filter moves across the input at each step, controlling how densely the image is sampled and how quickly spatial resolution is reduced. Filter sizes (11, 5, 3) are depicted as cubes and define the local receptive field of each convolution. The two horizontal streams reflect the model’s original two-GPU training setup, where feature maps were split across devices (e.g., 48 + 48 = 96 channels). The fully connected layers (4096 neurons, implemented as 2048 + 2048 across GPUs) feed into a 1000-way output for ImageNet classification. Used with permission granted by the lead author (Alex Krizhevsky).

Trace how pooling layers reduce spatial resolution through AlexNet’s architecture. section 2.5.1

How Sutskever's List covers it

How deep it goes

Defines local maximum selection and explains its computational and positional-robustness effects, without detailing pooling variants.

Examples

Uses AlexNet’s 3 × 3 windows, stride 2, and progressively reduced spatial dimensions as concrete examples.

What you can do after

Understand how max pooling preserves strong local activations while reducing spatial dimensions and computation.

How it connects

Connects max pooling to convolutional layers, feature maps, and spatial downsampling.

liveBookmanning.com

Read Sutskever's List on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  Math and Architectures of Deep Learning
• 

  liveBookmanning.com
  Deep Learning with R, Third EditionA max-pooling layer down-samples local regions by retaining their maximum values.
• 

  liveBookmanning.com
  Deep Learning with Python, Third EditionA two-dimensional pooling layer that downsamples feature maps by taking local maxima.
• 

  liveBookmanning.com
  Deep Learning with PyTorch, Second Edition
• 

  liveBookmanning.com
  Build a Text-to-Image Generator (from Scratch)Max pooling is a spatial downsampling operation that retains the maximum activation within each local window.
• 

  liveBookmanning.com
  Graph Neural Networks in Actionmax_pool is a PyTorch Geometric pooling function that coarsens a graph by pooling clustered nodes using maximum-valued features.
• 

  liveBookmanning.com
  Natural Language Processing in Action, Second EditionMax pooling is a downsampling operation that retains the largest activation within each local region of a feature map.
• 

  liveBookmanning.com
  Evolutionary Deep Learning: Genetic algorithms and neural networksMax pooling is a neural-network operation that downsamples local activation regions by retaining their maximum values.
• 

  liveBookmanning.com
  Machine Learning Algorithms in DepthMax pooling is a downsampling operation that retains the largest value within each local window of a feature sequence.

Compare how 10 books explain max pooling →

Related concepts

• downsamplingkind of · Downsampling is the reduction of data quantity, spatial resolution, temporal resolution, or sampling rate by retaining, selecting, or aggregating fewer representative values or observations.
• poolingkind of · Pooling is an operation that aggregates values or representations across regions or axes to produce a reduced or summarized representation.
• convolutional neural networkpart of · A convolutional neural network is a neural-network architecture that applies learned convolutional operations to extract patterns from structured data.
• strided convolutioncontrasts with · A strided convolution is a convolution whose filter moves across input positions by more than one step, typically reducing spatial or temporal resolution.
• convolutional layeruses · A convolutional layer is a neural-network layer that applies shared learned filters to local regions of structured input to produce feature representations.
• neural network layerkind of · A neural network layer is a computational stage or component that transforms input representations into outputs within a neural network.

This explainer was generated by Manning's knowledge graph from the text of Sutskever's List. The section links point to the parts of the book it draws on.