1 paper · 1 filter
Ashish Jha, Anh huy Phan, Razan Dibo +1
Training modern neural networks on large datasets is computationally and environmentally costly. We introduce GRAFT, a scalable in-training subset selection method that (i) extract…