Department of Applied Statistics at Yonsei University.
TITLE
Theoretical Advances in Efficient Machine Learning
ABSTRACT
Recent machine learning models are significantly large in terms of the number of parameters employed, and addressing the challenge of efficiently training and utilizing such extensive models is a critical research problem. In this talk, I will introduce recent theoretical works that contribute to resolving this issue. The topics include (1) pruning randomly weighted neural networks, (2) implementing programmable computers using transformers, and (3) efficient fine-tuning of large pre-trained models.