MENTAL MODEL · M6288

Pooling

Learn & MetacognitionHigh supportLearning Science

Included

Version 1.0.0 · Updated 2026-07-30

CORE DEFINITION

Downsampling feature maps (e.g., taking regional max or average) to reduce data volume, lower computational cost, and improve the model's tolerance to positional changes (invariance).

SCAFFOLDING EFFECT

psychology

Reduce cognitive load

Focus on the big picture and let go of details. When processing massive information (e.g., daily public opinion, user feedback), do not stare at every pixel. Perform "max pooling" (only look at the strongest feedback) or "average pooling" (look at overall trends), ignore noise, and capture key features.

anchor

Anchor fast decisions

On feature maps, use non-overlapping or sliding windows to take the max or average, reducing spatial resolution: decreasing parameter count and computation, suppressing overfitting, and providing translation invariance (small shifts do not change the pooling result). Essentially, it is "downsampling to capture the main structure and discard detailed noise."

MINIMUM ACTION

In progress 0/2

Practice this model in one real situation:

In CNNs, apply 2×2 or 3×3 pooling (with matching stride) after convolution
choose max pooling (to capture the strongest response) or average pooling (to capture overall trends) based on the task. Mapping to business: for massive feedback, use "max pooling" to see the strongest signals and "average pooling" to see overall trends.

Check to track your progress (stored locally)

Learning progress0%

account_treeGenealogyexpand_more

menu_bookReferencesexpand_more

Source support: Explicit

link
en.wikipedia.orghttps://en.wikipedia.org/wiki/PoolZH · Explicit
verified

RELATED MODELS