SNU Professor U Kang of the Department of Computer Science and Engineering has Research Paper Published in Neural Information Processing System

-Developed the KegNet method for compression of Deep learning Model into a faster and more precise model


<SNU Professor U Kang of Department of Computer Science and Engineering>

 
SNU College of Engineering announced on the 18th of September that Professor U Kang of the Department of Computer Science and Engineering and his team had their research paper on lightweight deep learning models using the Generative Adversarial Network (GAN) being published in Neural Information Processing System (NERIPS), an internationally renowned Artificial Intelligence Conference.
 
NeurIPS is the world's best academic society in the field of artificial intelligence. In his paper, Professor Kang proposed the KegNet technique of compressing complex deep learning models into faster, more efficient and concise models.
 
"Generally, a large amount of data is required to compress deep learning models, but it is difficult to use pre-existing techniques to compress models if only the pre-learned models are given without the provision of any data," Professor Kang said. "To overcome this, we used GAN from a given model to learn the data generator and proposed an effective model compression techniques through this," he added.

 

‘Key information regarding the lightweight deep learning model using GAN
(Generative Adversarial Network)’

 
The Research Paper is planned to be announced at the NeurIPS 2019 that is to be held in December 2019.
 
Jaemin Yoo, Minyong Cho, Taebum Kim, and U Kang, Knowledge Extraction
with No Observable Data, NeurIPS 2019, Vancouver, Canada.