overfit
Model memorizes training
To overfit is to tailor something so precisely to specific data or circumstances that it fails to generalize.
The term comes from statistics and machine learning, where a model that overfits captures noise and quirks in training data rather than underlying patterns: The model overfits—it performs perfectly on training data but poorly on new inputs.
The concept extends to any situation where excessive customization backfires: Don't overfit your resume to one job posting.
Deriving principles from examples
- regularizationPenalty preventing model complexity
- cross-validationResampling method for model assessment
- training setData used to teach models
- dropoutOne who quits early
underfitting · learning curve · pruning · ensemble · bagging · boosting · holdout · sample size
Memorization through repetition
- training dataInput for machine learning
- verbatimIn identical wording
- crammingIntensive last-minute study
- regurgitationEjecting swallowed material
Prefer simplest sufficient explanation
- nonlinearNot in sequence
- parametersDefined boundaries or constraints
- noiseUnwanted auditory disturbance
- parsimonyExcessive frugality with resources
generalization
Deriving principles from examples
- regularizationPenalty preventing model complexity
- cross-validationResampling method for model assessment
- training setData used to teach models
dropout · underfitting · learning curve · pruning · ensemble · bagging · boosting · holdout · sample size · capacity
rote learning
Memorization through repetition
- training dataInput for machine learning
- verbatimIn identical wording
- crammingIntensive last-minute study
Occam's razor
Prefer simplest sufficient explanation
- nonlinearNot in sequence
- parametersDefined boundaries or constraints
- noiseUnwanted auditory disturbance
overfitgeneralize
Make broadly applicable
generalization · robust · regularized · well-fitted · parsimonious
overfitunderfitting
Model too simple
oversimplified · underspecified · underpowered · parsimonious