Lottery Ticket Hypothesis: Finding Sparse, Trainable Neural Networks

Abstract

Lottery ticket hypothesis is one of the latest and most significant advances in the field of theoretical deep learning, suggesting that sparse yet effective representations can be found inside large overparametrized neural networks. We present a brief introduction to the context of neural network pruning and proceed to present three significant papers discussing the impact of lottery ticket hypothesis in deep learning research.

Date
Jul 18, 2019
Event
Statistical Machine Learning Course Presentation
Location
Room 4B, Building H2bis, University of Trieste, Italy

Related