B0845
Title: Pruning deep neural networks for lottery tickets
Authors: Rebekka Burkholz - CISPA Helmholtz Center for Information Security (Germany) [presenting]
Abstract: Deep learning continues to impress with breakthroughs across disciplines but comes at severe computational and memory costs that limit global participation in the development of related technologies. Can some of these challenges be addressed by finding and training smaller models? The lottery ticket hypothesis has given hope that this question might be answered by pruning randomly initialized neural networks. A strong version of this hypothesis is proven in realistic settings. Inspired by the theory, a framework is created that allows us to identify the current limitations of state-of-the-art algorithms in finding extremely sparse lottery tickets and highlight some opportunities for future progress.
