AI ART Discover · Learn · Create
L AI Glossary Term • Term 117 of 224
Plain English Definition

What does “Label Smoothing” mean?

Core Definition

A regularization method that softens hard labels to improve generalization.

Operational Mechanism

How Label Smoothing Works

Label Smoothing comparing Hard labels (cat = 1.0, dog = 0.0, bird = 0.0) with Smoothed labels (cat = 0.9, dog = 0.05, bird = 0.05).

Real-World Scenario

Practical Example

Instead of telling a classifier the right class has probability 1 and every other class 0, training uses 0.9 and spreads 0.1 among the others. This stops the model from becoming extremely overconfident.

Core Significance

Why It Matters

Smoothing is a simple regularizer that can improve generalization and calibration. It is a small change with often useful effects.

Expert Q&A

Common Questions About Label Smoothing

Q: Does it always improve accuracy?

Not always, but it often helps slightly, especially in large classifiers.

Q: Does it change what the model predicts?

It changes training targets, so the final probabilities are softer and the model is less extreme about its confidence.

Q: Are there downsides?

It can reduce the usefulness of the model's outputs for distillation in some cases, because softened probabilities carry less information.