Current Projects

Core machine intelligence algorithms

We primarily focus on overcoming the limitations of the current supervised machine learning framework. The first category we focus on is the adaptive machine learning (AML) schemes that deal with more challenging scenarios without large-scale labelled training data. The examples of those schemes include continual learning, meta/transfer learning, and self-supervised/unsupervised learning. The second focus category is the trustworthy machine learning (TML) methods that make ML models more reliable and safe to use. Namely, we pursue to make ML algorithms more explainable, fair, bias-free, and robust.

Some more concrete funded project regarding above topics includes the following:

  • Next generation continual learning (Supported by NRF (2021-2026), IITP (2022-2027), US AFOSR (2023-2025) )
  • XAI with causality / Robust XAI / Few-shot learning (Supported by IITP (2022-2027))
  • Fairness in machine learning (Supported by IITP (2019-2022))
  • Hyperscale AI / Multimodal learning (Supported by Naver (2021-2024))
  • Fair continual learning (Supported by LG AI Research (2022-2023))
  • Continual reinforcement learning (Supported by SAIT (2022-2025))

More on core algorithm research agenda is here.


Data science applications

For applications, we aim to apply the ML algorithms to solve a real-world problems that can lead to a high impact. Some of the recent topics include multi-modal model inference, neuroscience / medical image analyses, semiconductor random variation prediction, image restoration, radar-based activity recognition, satellite-based PM2.5 estimation, etc.

Some recent funded projects include the following:

  • Deep learning based battery management (Supported by Hyundai Motors) (2023-2024)
  • Neuroscience / Medical data analyses / Brain decoding (Supported by NRF (2021-2023, 2023-2028))


Past Projects

Following are some past projects.

  • Pre-training large-scale multi-modal model (Supported by IITP-MSRA (2021-2022))
  • Semiconductor random variation prediction (Supported by KEIT (2019-2021))
  • Image restoration (Supported by NRF, Samsung (2016-2019, 2022-2023))
  • Interpretable machine learning (Supported by KIST (2018-2020))
  • Adaptive machine learning for digital companion (Supported by IITP (2016-2020))
  • Satellite data based PM2.5 level estimation (Joint work with Prof. Yang Liu)
  • Neural network based denoising / estimation (Supported by NRF(2016-2019), Samsung(2018-2019))
  • Non-intrusive load monitoring (Jointly with Encored Technologies, Inc., SNU ADSL Lab)
  • DNA sequence denoising (Jointly with SNU DSAI Lab)
  • Doppler-Radar based Human Activity Recognition (Jointly with Prof. Youngwook Kim)