Distilled Pruning: Using Synthetic Data to Win the Lottery

Luke McDermott, Daniel Cummings

This work introduces a novel approach to pruning deep learning models by using distilled data. Unlike conventional strategies which primarily focus on architectural or algorithmic…