Machine Learning · Level 2 of 5
Overfitting
Learning training-specific patterns that fail to transfer well.
A model can fit noise as though it were a dependable signal.
Listen to the definition and example
Audio transcript
Overfitting. Learning training-specific patterns that fail to transfer well. A model can fit noise as though it were a dependable signal. For example: Training accuracy rises while validation accuracy falls.
Explore this concept
A useful analogy
Memorizing a practice exam’s quirks instead of learning the subject.
Why it matters
This helps you choose and evaluate a learning setup that matches the data and task.
Technical detail
A widening gap between training and representative validation performance can indicate fitting sample-specific variation. Diagnosis depends on the evaluation protocol.
Common misconception
More training accuracy is not automatically better generalization.
Start with
Related concepts
Quick recall question
Try answering before looking back at the definition.