Kernel size

As explained in Learn Generative AI with PyTorch

The spatial size of a convolutional kernel or receptive window applied to an input.

Kernel size is the spatial size of the filter, or receptive window, that a convolutional layer moves across an input image. In PyTorch, kernel_size=3 creates a 3 × 3 filter, while kernel_size=2 creates a 2 × 2 filter. Smaller windows inspect local regions; the selected size helps determine which spatial patterns the layer can extract. section 4.2.1 section 4.2

Read kernel size alongside the settings that control how a filter scans the image:

• Stride specifies how many pixels the filter moves horizontally or vertically after each operation; larger strides downsample more. section 4.2.1
• Padding adds rows of zeros around the input; its default is zero and it changes how the filter fits near image boundaries. section 4.2.1
• Kernel size differs from transposed convolution: convolution extracts features, whereas transposed convolution upsamples feature maps. section 4.0 section 4.2

A 2 × 2 convolution

A 3 × 3 grayscale image can be passed to nn.Conv2d with in_channels=1, out_channels=1, kernel_size=2, and stride=1. Replacing the learned weights with [[1, 2], [3, 4]] and setting the bias to 0 produces a 2 × 2 output tensor: [[7, 14], [54, 50]]. The output is smaller because the 2 × 2 window moves across the 3 × 3 input without padding. section 4.2.1

Figure 4.5 A numerical example of how convolutional operations work, with stride equal to 1 and no padding

Use the numerical figure to trace a 2 × 2 filter across a 3 × 3 input with stride 1 and no padding. section 4.2.1

How Learn Generative AI with PyTorch covers it

How deep it goes

Defines kernel size, relates it to stride and padding, and works through a fixed-weight 2 × 2 convolution without deriving output formulas.

Examples

Demonstrates a 2 × 2 kernel scanning a 3 × 3 grayscale tensor with stride 1, zero padding, fixed weights, and zero bias.

What it assumes

Assumes basic image tensors and CNN terminology while explaining channels, filters, kernels, stride, and padding directly.

What you can do after

Build a PyTorch Conv2d layer by selecting kernel_size, stride, channels, padding, and optional bias.

How it connects

Connects kernel size with convolutional layers, filters, stride, padding, downsampling, feature extraction, and transposed convolution.

liveBookmanning.com

Read Learn Generative AI with PyTorch on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  Evolutionary Deep Learning: Genetic algorithms and neural networksKernel size is the spatial extent of a convolutional filter applied by a convolutional layer.
• 

  liveBookmanning.com
  Math and Architectures of Deep LearningThe spatial extent of the sliding window used by a pooling layer, determining the patch size summarized at each step.
• 

  liveBookmanning.com
  Inside Deep Learning: Math, Algorithms, ModelsThe spatial extent of a convolutional filter.

Compare how 4 books explain kernel size →

Related concepts

• 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.
• max poolingused by · Max pooling is an aggregation operation that replaces each local or grouped region with its maximum value, often to downsample a representation.
• poolingused by · Pooling is an operation that aggregates values or representations across regions or axes to produce a reduced or summarized representation.
• hyperparameterkind of · A hyperparameter is a setting chosen outside ordinary parameter learning that controls a model, algorithm, or learning procedure.
• paddingcontrasts with · Padding is the addition of extra space, values, or data around or within a structure to satisfy size, alignment, layout, or processing requirements.
• expressive powercontrasts with · Expressive power is the capacity of a model, language, or system to represent or convey a range and complexity of functions, patterns, distinctions, or ideas.

This explainer was generated by Manning's knowledge graph from the text of Learn Generative AI with PyTorch. The section links point to the parts of the book it draws on.