Cognitive Scaffold

Preparing your thinking workspace

arrow_back_ios_new
MENTAL MODEL · M12890

Validation Set

Validation Set
Learn & MetacognitionHigh supportLearning Science
Included
account_tree

Version 1.0.0 · Updated 2026-07-30

CORE DEFINITION

Data split: training set / validation set / test set. Purpose: hyperparameter tuning and model selection, not used for final evaluation. Prevent overfitting: the model cannot 'see' the data used for final evaluation.

SCAFFOLDING EFFECT

psychology

Reduce cognitive load

Ensure the model's generalization ability. Performing well on 'known data' doesn't count; it must also perform well on 'unseen data'.

anchor

Anchor fast decisions

The validation set is a subset of data used during training for hyperparameter tuning, model selection, and early stopping, separated from the training set to avoid test set leakage. It estimates the model's performance on data that has been 'seen' but not used for training.

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

  • link
    zh.wikipedia.orghttps://zh.wikipedia.org/wiki/%E8%AE%AD%E7%BB%83%E9%9B%86%E3%80%81%E9%AA%8C%E8%AF%81%E9%9B%86%E5%92%8C%E6%B5%8B%E8%AF%95%E9%9B%86ZH · Explicit
    verified

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

Connectivism

Connectivism is a theoretical framework for understanding learning in a digital age. It emphasizes how internet technologies such as web browsers, search engines, wikis, online discussion forums, and

arrow_forward