ResNet
ResNet is a family of deep neural network architectures built around residual connections that help information and gradients propagate through the network.
Sutskever's List
ResNet is a deep neural network architecture that uses residual connections to facilitate the training of very deep networks.
- How it is used
- Uses residual blocks to make deep convolutional networks trainable and serve as backbones for computer-vision systems.
- How deep it goes
- Explains two-path blocks, gradient flow, bottleneck variants and v2 pre-activation through diagrams, comparisons and ablations rather than a full implementation.
- Examples
- Compares plain networks on CIFAR-10, ResNet variants on ImageNet, pre-activation ablations and a Faster R-CNN backbone replacement.
- What you can do after
- Evaluate/understand why skip paths improve deep-network optimization and why preserving an unaltered shortcut matters in very deep variants.
Inside Deep Learning: Math, Algorithms, Models
A neural network architecture built from residual blocks with skip connections that help information and gradients propagate through deep networks.
- How it is used
- Deploys residual blocks for deep image classification and adapts pretrained ResNet models to new classification targets.
- How deep it goes
- Implements bottleneck paths, shortcut projections, block stacking, pooling, and classifier replacement without deriving residual optimization mathematically.
- Examples
- Uses Fashion-MNIST residual classifiers, cats-versus-dogs ResNet-18 adaptation, ImageNet pretraining, and filter visualizations.
- What you can do after
- Apply it via PyTorch by composing residual modules, training a classifier, or replacing torchvision ResNet's final fully connected layer.
Read how Inside Deep Learning: Math, Algorithms, Models explains ResNet →
Math and Architectures of Deep Learning
A deep convolutional neural network architecture that uses identity shortcut connections to ease optimization of very deep networks.
- How it is used
- Explains how shortcut connections enable training much deeper convolutional networks when ordinary depth causes degradation.
- How deep it goes
- Connects degradation to identity-function learning, then descends into residual blocks, downsampling, and PyTorch ResNet-34 construction.
- Examples
- Uses CIFAR-10 error comparisons, residual-block diagrams, PyTorch BasicBlock code, and a ResNet-34 implementation.
- What you can do after
- Builds residual blocks and a ResNet-34-style model with PyTorch modules and downsampling shortcuts.
Read how Math and Architectures of Deep Learning explains ResNet →
Deep Learning with PyTorch, Second Edition
A family of residual neural network architectures that use skip connections to ease training of deep models.
- How it is used
- Uses skip connections to make very deep convolutional networks easier to train.
- How deep it goes
- Defines identity mappings, gradient shortcuts, and PyTorch blocks, but stops short of detailed residual-network variants.
- Examples
- Demonstrates NetDepth, NetRes, a reusable ResBlock, and NetResDeep with 100 stacked residual blocks.
- What you can do after
- Build it from scratch using PyTorch modules, convolutional layers, residual blocks, and skip connections.
Read how Deep Learning with PyTorch, Second Edition explains ResNet →
Also covered in
- Deep Learning with R, Third Edition
- Build a Text-to-Image Generator (from Scratch)
- Transfer Learning for Natural Language Processing
- Machine Learning Algorithms in Depth
What to know first
- linear layerA linear layer is a neural-network layer that applies a learned affine transformation to its input, typically using weights and a bias.
- batch normalizationBatch normalization is a neural-network technique that normalizes activations using statistics computed across a batch during training and stored running statistics during inference.
- bounding boxA bounding box is a coordinate-defined rectangular region in two or more dimensions that encloses or localizes an object, area, or set of points.
- convolutional layerA convolutional layer is a neural-network layer that applies shared learned filters to local regions of structured input to produce feature representations.
Related concepts
- residual connectionuses · A residual connection is a neural-network pathway that carries an earlier representation around one or more transformations so it can be combined with a later computation.
- ResNet50includes · ResNet50 is a 50-layer residual convolutional neural network architecture commonly used for image recognition and visual feature extraction.
- vanishing gradient problemaddresses · The vanishing-gradient problem is a neural-network training difficulty in which gradients shrink during backpropagation until parameter updates in earlier layers or steps become ineffective.
- vision modelkind of · A vision model is a computational or machine-learning model that processes visual data, such as images or video, to analyze, interpret, generate, or extract information.
- Faster R-CNNused by · Faster R-CNN is an object-detection architecture that uses a learned region proposal network and detection heads to localize and classify objects in images.
Questions about ResNet
What is ResNet?
ResNet is a family of deep neural network architectures built around residual connections that help information and gradients propagate through the network.
Which Manning books explain ResNet?
Sutskever's List, Inside Deep Learning: Math, Algorithms, Models, Math and Architectures of Deep Learning and Deep Learning with PyTorch, Second Edition.
What should I know before learning ResNet?
Manning books that teach ResNet usually introduce linear layer, batch normalization, bounding box and convolutional layer first.
What kinds of ResNet are there?
Kinds of ResNet include ResNet50.
What is ResNet a kind of?
ResNet is a kind of vision model. A vision model is a computational or machine-learning model that processes visual data, such as images or video, to analyze, interpret, generate, or extract information.
What is ResNet part of?
ResNet is part of CS231n. CS231n is a Stanford course on convolutional neural networks and deep learning for visual recognition.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.



