MENTAL MODEL · M0545

Backpropagation

TechnicalHigh supportNeuroscience

Included

Version 1.0.0 · Updated 2026-07-30

CORE DEFINITION

The core algorithm of neural networks. It calculates the error between the output and the true result, then propagates the error backward from the output layer to the input layer, adjusting the weights of neurons layer by layer to correct the model. Scaffold role: rapid iteration based on errors. It provides a high-level metaphor for learning: don't just look forward (doing problems), spend more time looking backward (correcting errors). Inferring which part's 'weight' went wrong from the erroneous result is the fastest path to improving learning efficiency.

SCAFFOLDING EFFECT

psychology

Reduce cognitive load

Rapid iteration based on errors. It provides a high-level metaphor for learning: don't just look forward (doing problems), spend more time looking backward (correcting errors). Inferring which part's 'weight' went wrong from the erroneous result is the fastest path to improving learning efficiency.

anchor

Anchor fast decisions

Neural network training algorithm: uses the chain rule to compute the gradient of the loss with respect to each parameter from the output layer backward to the input layer, then updates weights using gradient descent. It is the cornerstone of deep learning.

MINIMUM ACTION

In progress 0/4

Practice this model in one real situation:

Forward pass to get predictions and loss
backward pass to compute gradients layer by layer (chain rule)
update parameters according to learning rate
iterate until convergence and prevent gradient vanishing/exploding.

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/BackpropagationZH · Explicit
verified

RELATED MODELS