Caesar AI Atlas

Gradient Clipping

Caesar AI Atlas Definition

Gradient clipping is a technique that limits gradient values during training to prevent excessively large parameter updates. It is commonly used to mitigate the exploding gradient problem in deep and recurrent neural networks.

Other Definitions

Gradient Clipping Source

A commonly used mechanism to mitigate the exploding gradient problem by artificially limiting (clipping) the maximum value of gradients when using gradient descent to train a model.

Related Terms