Rui Li

I am a PhD student at Aalto University, supervised by Prof. Arno Solin. My research interests include Bayesian deep learning, uncertainty quantification, probabilistic machine learning and few-shot learning.

Previously, I obtained my Master degree in Machine Learning from University College London and Bachelor degree in Physics from Sun Yat-sen University.

Email: rui.li[at]aalto.fi

Google Scholar  /  GitHub  /  LinkedIn  /  CV

profile photo

News

  • 2025-04. Released the SUQ library for streamlined uncertainty quantification.
  • 2025-01. One paper accepted at ICLR 2025!
  • 2024-10. Three workshop papers accepted at NeurIPS 2024!

Selected Publications

See the full list of publications in my CV.

project image

Streamlining Prediction in Bayesian Deep Learning, ICLR 2025


Rui Li, Marcus Klasson, Arno Solin, Martin Trapp
paper / code

While estimating posterior has been actively researched in Bayesian deep learning (BDL), how to make predictions with posterior efficiently is largely overlooked. We examine streamlining prediction in BDL through a single forward pass without sampling. We showcase our approach for both MLP and transformers, such as ViT and GPT-2.
project image

Post-hoc Probabilistic Vision-Language Models, Under Review


Anton Baumann, Rui Li, Marcus Klasson, Santeri Mentu, Shyamgopal Karthik, Zeynep Akata, Arno Solin, Martin Trapp
paper / code

While Vision-Language models have shown remarkable performance in various tasks, the lack of uncertainty estimation makes them unreliable in high-stakes applications. We propose a post-hoc uncertainty quantification method based on Laplace approximation, which provides useful predictive uncertainties and better calibration.
project image

Improving Hyperparameter Learning under Approximate Inference in Gaussian Process Models, ICML 2023


Rui Li, Ti John, Arno Solin
paper / code

Variational inference and expectation propagation are two commonly used approximate inferences in Gaussian process models with complementary advantages. We developed a hybrid training procedure to bring the best of both worlds.
project image

Flatness Improves Backbone Generalisation in Few-shot Classification, WACV 2025 (Oral)


Rui Li, Martin Trapp, Marcus Klasson, Arno Solin
paper / code

In few-shot classification most efforts focus on adapting the backbone to the target domain without considering the importance of backbone training. We show flatness-aware backbone training can lead to better generalisation through theoretical and empirical results.




Design and source code from Jon Barron's website