What is overfitting in deep learning?

Overfitting happens when a deep-learning model learns the training data too closely and performs well on that data but poorly on new, unseen data.

A typical sign is:

Training loss   → keeps decreasing
Validation loss → starts increasing

This means the model is improving at fitting the training set but becoming worse at generalizing.

Why does overfitting happen?

Common causes include:

• The model is too complex for the amount of available data.
• Training continues for too many epochs.
• The dataset is too small or not diverse enough.
• The model starts fitting noise or accidental patterns instead of generalizable structure.

Simple example

Suppose a model reaches:

Training accuracy   = 99%
Validation accuracy = 82%

The large gap suggests the model may be overfitting.

The goal is not simply to perform well on the training set, but to perform well on data the model has not seen before.

Techniques used to reduce overfitting include regularization, dropout, data augmentation, and early stopping. These are covered in their own dedicated answers.

In short, overfitting means a model fits the training data very well but fails to generalize equally well to unseen data.