Minyoung Kim

• Homepage: https://minyoungkim21.github.io/
• Github repos: https://github.com/minyoungkim21
• Google scholar: https://scholar.google.com/citations?hl=en&user=tKpOSMsAAAAJ&view_op=list_works&sortby=pubdate
• CV: Upon request

Recent Positions

• Staff Research Scientist at Samsung AI Centre, Cambridge, UK (Dec. 2022 ~ Present)
• Senior Research Scientist at Samsung AI Centre, Cambridge, UK (Mar. 2019 ~ Nov. 2022)

Research Interest

• General deep learning
  • LLM training and sparse fine-tuning
  • Transformer architecture analysis
  • Meta learning and few-shot learning
• Bayesian statistics
  • Graphical models
  • Bayesian deep learning
  • Gaussian processes and neural processes
• Generative models
  • Diffusion models for conditional/unconditional generation
  • Diffusion models for bridging and flow matching
  • Stochastic event models (Poisson/Cox processes)

Publications (Recent 6 years)

(For the full list of papers, please refer to my Google Scholar page)

1. LIFT: Learning to Fine-tune via Bayesian Parameter Efficient Meta Fine-tuning
   Minyoung Kim, Timothy Hospedales
   International Conference on Learning Representations (ICLR), 2025 (Spotlight)

2. Model Diffusion for Certifiable Few-shot Transfer Learning
   Fady Rezk, Royson Lee, Henry Gouk, Timothy Hospedales, Minyoung Kim
   arXiv preprint arXiv:2502.06970, 2025

3. FedP$^2$EFT: Federated Learning to Personalize Parameter Efficient Fine-Tuning for Multilingual LLMs
   Royson Lee, Minyoung Kim, Fady Rezk, Rui Li, Stylianos I. Venieris, Timothy Hospedales
   arXiv preprint arXiv:2502.04387, 2025

4. A Stochastic Approach to Bi-Level Optimization for Hyperparameter Optimization and Meta Learning
   Minyoung Kim, Timothy Hospedales
   AAAI Conference, Philadelphia, PA, 2025

5. A Bayesian Approach to Data Point Selection
   Xinnou Xu, Minyoung Kim, Royson Lee, Timothy Hospedales, Brais Martinez
   (The first two authors equal contribution)
   Advances in Neural Information Processing Systems (NeurIPS), 2024

6. A Hierarchical Bayesian Model for Few-Shot Meta Learning
   Minyoung Kim, Timothy Hospedales
   International Conference on Learning Representations (ICLR), 2024 (Spotlight)

7. FedL2P: Federated Learning to Personalize
   Royson Lee, Minyoung Kim, Da Li, Xinchi Qiu, Timothy Hospedales, Ferenc Huszar, Nicholas Lane
   Code
   Advances in Neural Information Processing Systems (NeurIPS), 2023

8. BayesTune: Bayesian Sparse Deep Model Fine-tuning
   Minyoung Kim, Timothy Hospedales
   Code
   Advances in Neural Information Processing Systems (NeurIPS), 2023

9. BayesDLL: Bayesian Deep Learning Library
   Minyoung Kim, Timothy Hospedales
   Code
   arXiv preprint arXiv:2309.12928, 2023

10. Domain Generalisation via Domain Adaptation: An Adversarial Fourier Amplitude Approach
    Minyoung Kim, Da Li, Timothy Hospedales
    International Conference on Learning Representations (ICLR), 2023

11. SwAMP: Swapped assignment of multi-modal pairs for cross-modal retrieval
    Minyoung Kim
    Artificial Intelligence and Statistics (AISTATS), 2023

12. FedHB: Hierarchical Bayesian Federated Learning
    Minyoung Kim, Timothy Hospedales
    arXiv preprint arXiv:2305.04979, 2023

13. Fisher SAM: Information Geometry and Sharpness Aware Minimisation
    Minyoung Kim, Da Li, Shell Xu Hu, Timothy M. Hospedales
    International Conference on Machine Learning (ICML), 2022

14. Pushing the Limits of Simple Pipelines for Practical Few-Shot Learning
    Shell Xu Hu, Da Li, Jan Stuhmer, Minyoung Kim, Timothy M. Hospedales
    Code
    IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2022

15. Gaussian Process Modeling of Approximate Inference Errors for Variational Autoencoders
    Minyoung Kim
    IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2022

16. Differentiable Expectation-Maximization for Set Representation Learning
    Minyoung Kim
    International Conference on Learning Representations (ICLR), 2022

17. Variational Continual Proxy-Anchor for Deep Metric Learning
    Minyoung Kim, Ricardo Guerrero, Hai X. Pham, Vladimir Pavlovic
    Artificial Intelligence and Statistics (AISTATS), 2022

18. Gaussian Process Meta Few-shot Classifier Learning via Linear Discriminant Laplace Approximation
    Minyoung Kim, Timothy Hospedales
    arXiv preprint arXiv:2111.05392, 2021

19. On PyTorch Implementation of Density Estimators for von Mises-Fisher and Its Mixture
    Minyoung Kim
    Code
    arXiv preprint arXiv:2102.05340, 2021

20. Learning Disentangled Factors from Paired Data in Cross-Modal Retrieval: An Implicit Identifiable VAE Approach
    Minyoung Kim, Ricardo Guerrero, Vladimir Pavlovic
    ACM International Conference on Multimedia (MM), 2021

21. Recursive Inference for Variational Autoencoders
    Minyoung Kim and Vladimir Pavlovic
    Neural Information Processing Systems (NeurIPS), 2020

22. Ordinal-Content VAE: Isolating Ordinal-Valued Content Factors in Deep Latent Variable Models
    Minyoung Kim and Vladimir Pavlovic
    International Conference on Data Mining (ICDM), Sorrento, Italy, 2020

23. Bayes-Factor-VAE: Hierarchical Bayesian Deep Auto-Encoder Models for Factor Disentanglement
    Minyoung Kim, Yuting Wang, Pritish Sahu, Vladimir Pavlovic
    IEEE International Conference on Computer Vision (ICCV), Seoul, Korea, 2019 (Oral)

24. Efficient Deep Gaussian Process Models for Variable-Sized Inputs
    Issam H. Laradji, Mark Schmidt, Vladimir Pavlovic, Minyoung Kim
    International Joint Conference on Neural Networks (IJCNN), Budapest, Hungary, 2019

25. Relevance Factor VAE: Learning and Identifying Disentangled Factors
    Minyoung Kim, Yuting Wang, Pritish Sahu, Vladimir Pavlovic
    arXiv preprint arXiv:1902.01568, 2019

26. Sparse large-margin nearest neighbor embedding via greedy dyad functional optimization
    Minyoung Kim
    Applied Intelligence 49 (10), 3628-3640, 2019

27. Large-Margin Learning of Cox Proportional Hazard Models for Survival Analysis
    Minyoung Kim
    Applied Intelligence, 49 (5), 1675-1687, 2019

28. Unsupervised Visual Domain Adaptation: A Deep Max-Margin Gaussian Process Approach
    Minyoung Kim, Pritish Sahu, B. Gholami, Vladimir Pavlovic
    IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Long Beach, CA, USA, 2019 (Oral)

Patents

1. Method and system for federated learning (2024)
   Minyoung Kim, Timothy Hospedales
   US Patent App. 18/512,195
   (A novel hierarchical Bayesian model for effective and efficient Federated Learning)

2. Method and apparatus for generating a noise-resilient machine learning model (2024)
   Minyoung Kim, Timothy Hospedales, Da Li, Xu Hu
   EPO 4374297 - EP23747321A1
   (A new proposal of loss curvature for robust deep learning to incorporate underlying Fisher information geometry and smoothness constraints)

3. Method and apparatus for meta few-shot learner (2023)
   Minyoung Kim, Timothy Hospedales
   US Patent App. 17/843,590
   (A new hierarchical Bayesian approach to tackling multi-task few-shot meta learning)

4. Method and apparatus for concept matching (2023)
   Vladimir Pavlovic, Minyoung Kim, Ricardo Guerrero, Hai Pham
   US Patent App. 17/434,314
   (A novel shared latent space generative model approach for multi-modal data)

5. Object tracking in video with visual constraints (2013)
   Minyoung Kim, Sanjiv Kumar, Henry A. Rowley
   US Patent 8,477,998
   (An object/face tracking system that is robust to noise and occlusion via hidden-Markov Fisher-discriminant latent space state regularization)

Awards

1. Silver Medal in Samsung Best Paper Awards 2024
   BayesTune: Bayesian Sparse Deep Model Fine-tuning

2. Bronze Medal in Samsung Best Paper Awards 2023
   Swamp: Swapped assignment of multi-modal pairs for cross-modal retrieval

3. Merit Award in Samsung Best Paper Awards 2023
   Domain Generalization via Domain Adaptation: An Adversarial Fourier Amplitude Approach

4. Bronze Medal in Samsung Best Paper Awards 2022
   Gaussian Process Modeling of Approximate Inference Errors for VAE

5. Bronze Medal in Samsung Best Paper Awards 2021
   Recursive Inference for Variational Autoencoders

6. Bronze Medal in Samsung Best Paper Awards 2019
   Gaussian Process Domain Adaptation

Programming Skills

Python (PyTorch/JAX/TensorFlow); Matlab; Javascript; C/C++; Java; Perl