How deeply does Deep Learning with R, Third Edition cover convolutional neural network?

Works through model construction, training and evaluation, then explains local convolutions and feature maps without a mathematical derivation.

Convolutional neural network

As explained in Deep Learning with R, Third Edition

A convolutional neural network is a neural network that applies convolution-based operations to learn spatial or locally structured features, especially from grid-like data such as images.

A convolutional neural network (convnet or CNN) learns to recognize images from local patterns. Rather than connecting every pixel to a layer that learns patterns across the whole image, convolution layers look at small windows. A pattern learned in one location can then be recognized elsewhere. section 8.1.1

Classify handwritten digits

For 28 × 28 grayscale MNIST images, start with an input shape of (28, 28, 1). Stack Conv2D and MaxPooling2D layers, pool the final feature maps, and use a 10-unit softmax layer to predict the digit. Reshape images to include their channel and divide pixel values by 255 before training. This small model reaches 99.31% test accuracy in the reported run, compared with 97.8% for an earlier densely connected model. section 8.1

Listing 8.1 Instantiating a small convnet

library(keras3)

inputs <- keras_input(shape = c(28, 28, 1))
outputs <- inputs |>
  layer_conv_2d(filters = 64, kernel_size = 3, activation = "relu") |>
  layer_max_pooling_2d(pool_size = 2) |>
  layer_conv_2d(filters = 128, kernel_size = 3, activation = "relu") |>
  layer_max_pooling_2d(pool_size = 2) |>
  layer_conv_2d(filters = 256, kernel_size = 3, activation = "relu") |>
  layer_global_average_pooling_2d() |>
  layer_dense(10, activation = "softmax")

model <- keras_model(inputs = inputs, outputs = outputs)

The R/Keras model definition shows the full path from image input through alternating convolution and pooling layers to a softmax classifier. section 8.1

Each convolution applies the same learned transformation to patches of an input feature map. A feature map has height, width, and channels; its output channels indicate learned patterns rather than image colors. Early layers can detect edges and textures, while later layers combine simpler features into more abstract ones. As a typical convnet gets deeper, its feature maps become smaller across the image and deeper across channels. section 8.1.1 section 8.2.3

For a harder dogs-versus-cats task, a larger convnet starts with 180 × 180 color images and adds more convolution-and-pooling stages. It rescales pixels, then ends with one sigmoid output for binary classification. With only 2,000 training images, training from scratch establishes a useful baseline, but overfitting becomes the main obstacle. Data augmentation helps; feature extraction and fine-tuning of a pretrained convnet offer further options for small datasets. section 8.2 section 8.2.3

Listing 8.7 Instantiating a small convnet for dogs vs. cats classification

inputs <- keras_input(shape = c(180, 180, 3))  #1
outputs <- inputs |>
  layer_rescaling(1 / 255) |>  #2
  layer_conv_2d(filters = 32, kernel_size = 3, activation = "relu") |>
  layer_max_pooling_2d(pool_size = 2) |>
  layer_conv_2d(filters = 64, kernel_size = 3, activation = "relu") |>
  layer_max_pooling_2d(pool_size = 2) |>
  layer_conv_2d(filters = 128, kernel_size = 3, activation = "relu") |>
  layer_max_pooling_2d(pool_size = 2) |>
  layer_conv_2d(filters = 256, kernel_size = 3, activation = "relu") |>
  layer_max_pooling_2d(pool_size = 2) |>
  layer_conv_2d(filters = 512, kernel_size = 3, activation = "relu") |>
  layer_global_average_pooling_2d() |>  #3
  layer_dense(1, activation = "sigmoid")

model <- keras_model(inputs, outputs)

The larger R/Keras model makes the changes for color images and binary classification concrete. section 8.2.3

How Deep Learning with R, Third Edition covers it

Cover of Deep Learning with R, Third Edition

liveBook manning.com

Read Deep Learning with R, Third Edition on liveBook Get the book at manning.com

Also explained in

liveBook manning.com

Natural Language Processing in Action, Second EditionA convolutional neural network is a neural network that applies learned convolutional filters to structured inputs to detect local patterns and construct hierarchical representations.
liveBook manning.com

Sutskever's ListA convolutional neural network is a neural network architecture that applies learned convolutional filters to structured input such as images.
liveBook manning.com

Deep Learning with Python, Third EditionA neural network built from convolution and pooling layers that processes data with local spatial structure.
liveBook manning.com

Evolutionary Deep Learning: Genetic algorithms and neural networksA convolutional neural network is a neural network that processes structured inputs with learned convolutional filters, typically for tasks such as image classification.
liveBook manning.com

Deep Learning with PyTorch, Second EditionA neural network that uses convolutional layers to process spatial data.
liveBook manning.com

Distributed Machine Learning PatternsA convolutional neural network is a neural network model that uses convolutional operations to learn spatial or locally structured representations.
liveBook manning.com

Inside Deep Learning: Math, Algorithms, ModelsA neural-network design that uses convolutional layers to exploit spatial structure in image data.
liveBook manning.com

Graph Neural Networks in ActionA convolutional neural network is a neural network that applies learned convolutional filters to exploit local spatial structure, especially in grid-like data such as images.
liveBook manning.com

Learn Generative AI with PyTorchA neural network architecture using convolution operations, here applied to multidimensional music data to extract spatial features.
liveBook manning.com

The AI Pocket BookA neural network architecture that applies convolutions and downsampling to learn spatial features from grid-like data such as images.
liveBook manning.com

Data Without LabelsA convolutional neural network is a neural-network model that uses convolutional operations to learn spatial or locally structured features.
liveBook manning.com

Math and Architectures of Deep LearningA neural network that uses convolutional layers to process structured grid data such as images.
liveBook manning.com

Grokking Machine LearningA neural-network architecture that uses convolutional operations to learn localized and hierarchical patterns, especially in structured data such as images.
liveBook manning.com

Grokking AI Algorithms, Second EditionA convolutional neural network is a neural-network architecture that applies learned convolutional filters and typically pooling operations to detect spatial patterns in structured data such as images.
liveBook manning.com

Delivering Machine Learning Projects: From design to deploymentA convolutional network is a neural network architecture that applies local filters over structured inputs to capture spatial or local patterns.
liveBook manning.com

Machine Learning Algorithms in DepthA convolutional neural network is a neural-network architecture that uses convolution and pooling operations to process structured spatial or grid-like data.
liveBook manning.com

Inside AIA convolutional neural network is a neural-network model that applies learned convolutional filters to structured data such as images.

Compare how 18 books explain convolutional neural network →

Related concepts

language modelkind of · A language model is a computational model that represents patterns in linguistic tokens to estimate sequence probabilities and predict or generate language from context.
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.
poolinguses · Pooling is an operation that aggregates values or representations across regions or axes to produce a reduced or summarized representation.
convolution kerneluses · A convolution kernel is a small array or tensor of weights applied across local input regions to compute weighted responses and detect or transform patterns.
U-Netincludes · U-Net is an encoder-decoder neural-network architecture with downsampling and upsampling paths connected by skip connections for combining features at multiple spatial scales.
LeNetincludes · LeNet is an early convolutional neural-network architecture designed primarily for handwritten-digit and character recognition.

This explainer was generated by Manning's knowledge graph from the text of Deep Learning with R, Third Edition. The section links point to the parts of the book it draws on.