1 Hit in 6.1 sec

Lottery Tickets on a Data Diet: Finding Initializations with Sparse Trainable Networks [article]

Mansheej Paul, Brett W. Larsen, Surya Ganguli, Jonathan Frankle, Gintare Karolina Dziugaite
2022 arXiv   pre-print
network correlates well with good initializations for IMP.  ...  We additionally observe that by pre-training only on "easy" training data, we can decrease the number of steps necessary to find a good initialization for IMP compared to training on the full dataset or  ...  Roy for feedback on multiple drafts.  ... 
arXiv:2206.01278v1 fatcat:v2miqupbavbrharhw2qtkh7uam