Pruning

Pruning removes less important weights or neurons from a trained neural network, reducing model size and computation while aiming to preserve accuracy. It creates sparse models that are faster and more efficient for deployment.