Sungyoon Kim

This article discusses methods for training two-layer ReLU neural networks efficiently.

Introducing MoEfier for efficient transformation of language models with minimal training.

Explore the loss landscape and the role of regularization in neural networks.