Global Average Pooling

What is Global Average Pooling?

An operation that takes the average of the entire feature map, collapsing it to a single value per channel. Used to connect convolutional layers to the final classification layer.

Where did the term "Global Average Pooling" come from?

Replaces parameter-heavy fully connected layers.

How is "Global Average Pooling" used today?

Used in modern architectures (ResNet, EfficientNet).

Related Terms

• Pooling Layer
• Convolutional Neural Network (CNN)