Curriculum Vitae (Last updated on February 19, 2024)

I am Giung Nam, a Ph.D. student in the Kim Jaechul Graduate School of AI at Korea Advanced Institute of Science and Technology (KAIST AI), advised by Professor Juho Lee. I earned an M.S. degree in the KAIST AI and a B.S. degree in the Department of Computer Science and Engineering at Korea University.


Currently, my doctoral research centers on the topic of "Towards a scalable Bayesian deep learning." Here is a concise overview of my research interests:

  1. Modern deep neural networks are typically over-parameterized and thus under-specified by the available data. In such cases, numerous parameter configurations can equally effectively explain the data. Hence, it is more suitable to explore all potential hypotheses using Bayesian marginalization rather than seeking a single hypothesis solely through optimization.
  2. Although Bayesian marginalization is an attractive approach for large-scale deep neural networks, the primary challenge lies in its computational expense. To illustrate, computing the Bayesian model averaging integral through Monte Carlo integration requires multiple model copies (space complexity), as well as conducting forward passes for each of them (time complexity).
  3. Consequently, one of my main research directions involves exploring efficient approaches, both in terms of training and testing, to perform Bayesian inference in contemporary, large-scale deep neural network models. This area of research covers subfields associated with model compression, which includes techniques like pruning, distillation, and quantization.
  4. Moreover, it is now a common practice to adapt pre-trained models, such as foundation models, to particular tasks rather than starting model training from scratch. Given the valuable prior knowledge unquestionably contained within pre-trained models, another research direction involves finding effective methods for harnessing this knowledge through Bayesian principles.

IMPORTANT NOTE: Every masculine gender of the Republic of Korea shall faithfully perform military service, as prescribed by the Constitution of the Republic of Korea and the Military Service Act (Article 3(1) of the Military Service Act). Accordingly, I am set to start my role as a Technical Research Personnel on March 1, 2024. The Technical Research Personnel program offers an alternative way of fulfilling military service obligations, allowing designated research institutes (such as KAIST in my case) under the Military Manpower Administration to employ research personnel to contribute to advancing science and technology in Korea. Consequently, I expect to complete my doctoral degree by February 28, 2026.