LeNet

As explained in Math and Architectures of Deep Learning

A convolutional neural network architecture introduced for digit recognition and used here as the example image classifier.

LeNet is a convolutional neural network for classifying handwritten digits. It maps a centered image to one of ten labels, using convolution, pooling, and nonlinear activation to build increasingly useful visual representations. section 11.1

Its architecture follows a compact sequence of spatial feature extraction and dimensionality reduction:

• Resize each 28 × 28 MNIST image to the 32 × 32 input expected by LeNet. section 11.1
• Apply three convolutional layers with 5 × 5 kernels and stride 1, producing feature maps of 28 × 28, 10 × 10, and 1 × 1 with 6, 16, and 120 channels. section 11.1
• Follow the first two convolutional layers with 2 × 2 average-pooling layers, halving spatial resolution and reducing sensitivity to shifts and distortions. section 11.1
• Use tanh after each feature-map layer to introduce the nonlinearity needed to model nonlinear combinations of inputs. section 11.1

MNIST classifier in PyTorch Lightning

A practical implementation separates data preparation from model logic. An MNISTDataModule resizes images and provides loaders; a LeNetClassifier groups convolution, tanh, and average-pooling operations. A Trainer then fits the model, while prediction runs under torch.no_grad(). section 11.2.4

Figure 11.3 LeNet. (Source: “Gradient-based learning applied to document recognition”; http://mng.bz/Wz0a.)

Use the architecture figure to inspect LeNet’s layer sequence, feature-map sizes, pooling stages, and transition to the flattened representation. section 11.1

LeNet is effective for simple digit recognition but does not provide enough expressive power for complex real-world images. Deeper networks extend the same convolutional idea with more layers and nonlinear transformations. section 11.2 section 11.5

How Math and Architectures of Deep Learning covers it

How deep it goes

Explains layer dimensions, pooling, nonlinear activation, and implementation, but stops short of detailed optimization internals.

Examples

Uses MNIST digit images, LeNet layer dimensions, PyTorch Lightning modules, training, and inference as concrete examples.

What it assumes

Assumes familiarity with neural networks while revisiting convolution, pooling, and activation concepts through the MNIST task.

What you can do after

Build and train a LeNet digit classifier with PyTorch Lightning, then run inference on a test batch.

How it connects

Connects convolutional layers with pooling, tanh activation, MNIST, PyTorch, and deeper neural networks.

liveBookmanning.com

Read Math and Architectures of Deep Learning on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  Distributed Machine Learning PatternsLeNet is a convolutional neural network architecture originally designed for handwritten-digit recognition.
• 

  liveBookmanning.com
  Machine Learning Algorithms in DepthLeNet is a convolutional neural network architecture designed for visual recognition tasks.

Compare how 3 books explain LeNet →

Related concepts

• convolutional neural networkkind of · A convolutional neural network is a neural-network architecture that applies learned convolutional operations to extract patterns from structured data.
• MNIST datasetuses · MNIST is a labeled dataset of 28-by-28-pixel grayscale images representing handwritten digits from zero through nine.
• classifieraddresses · A classifier is a predictive computational model that assigns inputs to one or more discrete classes or categories.
• convolutional layerpart of · A convolutional layer is a neural-network layer that applies shared learned filters to local regions of structured input to produce feature representations.
• PyTorchuses · PyTorch is an open-source machine-learning framework for Python that provides tensor computation, automatic differentiation, and tools for building and training neural networks.
• 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.

This explainer was generated by Manning's knowledge graph from the text of Math and Architectures of Deep Learning. The section links point to the parts of the book it draws on.