Cognitive Scaffold

Preparing your thinking workspace

arrow_back_ios_new
MENTAL MODEL · M5920

Dropout

Dropout
TechnicalHigh supportNeuroscience
Included
account_tree

Version 1.0.0 · Updated 2026-07-30

CORE DEFINITION

In training neural networks, randomly "dropping out" (temporarily disabling) a portion of neurons. This forces the remaining neurons to learn more robust features, preventing the network from over-relying on specific pathways (overfitting).

SCAFFOLDING EFFECT

psychology

Reduce cognitive load

Becoming stronger through amputation. In team or personal growth, occasionally "disconnecting", forced "vacation", or "job rotation" (dropping some capabilities) can force the system to find new paths when resources are missing, greatly enhancing the system's resilience and adaptability.

anchor

Anchor fast decisions

In neural network training, randomly disabling some neurons forces the remaining neurons not to rely on specific pathways, learning more robust features and alleviating overfitting. The mechanism is: random "absence" breaks the co-adaptation among neurons, equivalent to training a slightly different subnetwork each iteration, and integrating them yields more robustness.

MINIMUM ACTION

In progress 0/4

Practice this model in one real situation:

Check to track your progress (stored locally)
Learning progress0%
account_treeGenealogyexpand_more
menu_bookReferencesexpand_more

Source support: Explicit

RELATED MODELS

Related model

Catastrophic Forgetting

Catastrophic interference, also known as catastrophic forgetting, is the tendency of an artificial neural network to abruptly and drastically forget previously learned information upon learning new information.

arrow_forwardRelated model

Grokking

A delayed-generalization phenomenon in machine learning: test performance can remain poor long after training data have been fitted, then improve substantially.

arrow_forwardRelated model

SCARF Model

Proposed by David Rock

arrow_forward