Pruning
Domain: MLOps
🎯 Motivation
Removing unimportant neurons or connections. Smaller model, minimal accuracy loss.
📋 Concrete Example
Pruning: remove 30% of neurons with the smallest weights. Model size: -30%. Speed: +40%. Accuracy: 88% -> 87%. Minimal loss for big gains.
🔗 Analogy
Trimming a tree - cut dead branches. The tree is smaller but healthier. Pruning removes unimportant parts of the model.