🏋️ Too much exercise 🤖 Too much learning

overtraining

6.6

Teaching a computer program too much with the same examples that it fails on new data

noun
/ˌoʊvərˈtreɪnɪŋ/

Specialized

: sobreajuste

If you do not use proper validation, your model may suffer from overtraining and fail on new data.

Examples

Regular testing on separate data sets helps avoid overtraining in machine learning projects.
The model suffered from overtraining after too many iterations.
The researchers adjusted their algorithm to prevent overtraining during the learning process.
Avoiding overtraining is crucial for model performance.
If you do not use proper validation, your model may suffer from overtraining and fail on new data.
Regular testing on separate data sets helps avoid overtraining in machine learning projects.
The researchers adjusted their algorithm to prevent overtraining during the learning process.

Synonyms

overfitting

vsovertraining

Specialized

8.1

A model that learns the training data too well and fails on new data

is caused specifically by training the model too much rather than other reasons

memorization

vsovertraining

Specialized

6.1

A model remembers examples it was trained on too exactly and fails on new data

is about that same failure but specifically from excessive training steps

How Extensive

overtraining

Specialized

6.6

retraining

Specialized

3.9

training

Specialized

85 2.8

pretraining

Specialized

4.4

Surface Forms

overtraining singular

overtrainings plural

Morphology

overtraining = training (semi-transparent) = over + train + ing

The morphological composition 'over-' + 'training' is clear, but the technical implication for machine-learning models (poor generalization to new data) requires domain knowledge beyond the base words.

Etymology

The word overtraining also uses over- 'too much' and train 'to teach', and in machine learning it means a model has been 'taught too much' on its training data so it then works poorly on new data.