Fetching the paper…

RED++ : Data-Free Pruning of Deep Neural Networks via Input Splitting and Output Merging · Around