Pruning

A model compression technique that removes unnecessary or low-importance parameters from a trained neural network to make it smaller and faster. Just as pruning a tree removes dead branches without harming its health, model pruning reduces complexity without significantly hurting performance.