You are reading immutable version 3. The current guide may be newer.

Study the paper

Deep Residual Learning for Image Recognition

Lessons, visuals, quizzes, flashcards, and resources—organized in teaching order.

12 activities

Lesson

At a glance

Introduction to Residual Learning

Introduction to Residual Learning

Lesson

Mathematical Formulation of ResNets

Mathematical Formulation of ResNets

A residual building block can be defined formally using shortcut connections. For a set of stacked layers that learn a residual mapping mathcal F ( mathbf x , W i ) , the output mathbf y is computed by adding the input mathbf x directly to the output of the stacked layers:

Lesson

Identity Shortcut Connection

Mathematical Formulation of ResNets

The fundamental formulation of deep residual learning introduces an identity shortcut connection. Instead of forcing stacked layers to directly fit a desired underlying mapping mathcal H ( mathbf x ) , we let these layers approximate a residual mapping mathcal F ( mathbf x , W i ) = mathcal H ( mathbf x ) - mathbf x . The original mapping is recast into mathcal F ( mathbf x , W i ) + mathbf x .

Lesson

Projection Shortcut Connection

Mathematical Formulation of ResNets

This equation defines the projection shortcut connection in Deep Residual Networks (ResNets). When the input dimension and output dimension of a residual block differ, a linear projection matrix W s is applied to the input mathbf x to match the dimensions of the residual function mathcal F ( mathbf x , W i ) .

Visual

Projection Shortcut Connection: step by step

Mathematical Formulation of ResNets

Follow the existing bounded walkthrough in its intended sequence.

Visual

Identity Shortcut Connection: step by step

Mathematical Formulation of ResNets

Follow the existing bounded walkthrough in its intended sequence.

Lesson

Understanding Bottleneck Architectures

Bottleneck Architectures for Deep ResNets

The Bottleneck Design in Deep ResNets

Lesson

Empirical Analysis and Extreme Depth Behavior

Empirical Analysis and Layer Responses

Empirical Analysis of Layer Responses

Lesson

Standard deviations of layer responses on CIFAR-10

Empirical Analysis and Layer Responses

The standard deviations (std) of layer responses on CIFAR-10 are analyzed to understand the behavior of residual learning. The responses are measured at the outputs of each 3 times3 layer, after Batch Normalization (BN) and before nonlinearity.

Quiz

Test your understanding

ResNet Knowledge Assessment

12 questions grounded in this paper section.

Resource

Further learning

ResNet Knowledge Assessment

3 supplementary resources for this paper section.

Resource

Research and implementation context

ResNet Knowledge Assessment

Explore notable related work and public implementation context.