Author

Me

Yuanbang Liang

PhD candidate in computer science @ Cardiff University

I'm working with Professor Yu-kun Lai at Cardiff University, as a research associate. My research interests are centred around machine learning and its applications in computer vision, computer graphics and content generation.

My PhD research, which is fully funded by the EPSRC, focuses on the hubness phenomenon to uncover the relationship between the hyper-dimensional distribution and the generative models. Now my research is on explainability of generative models to make model-generated results more aligned with objective facts is a core challenge in AI, often referred to as AI alignment. The goal is to ensure that an AI system's behaviour and decisions align with human intent, values, and objective reality.

Now I'm Expecting to graduate before 2025 and looking for the research jobs.

News

2025 June
2024 May

Our paper is accepted by ICML as Spotlight.

2024 Feb

Our paper is accepted by CVPR as Oral.

Education

2021.10 - till now

Cardiff University

Doctor of Philosophy

Thesis title: Uncovering the “Instincts” of Deep Generative Models for Fair and Unbiased Visual Content Creation. Supervised by Yipeng Qin, Yu-Kun Lai and Jing Wu.

Deep LearningGenerative ModelsHubness Phenomenon
2020.09 - 2021.09

University of Bristol

Master of Science

Engineering Mathematics.

Partial Differential EquationsApplied StatisticsMachine Learning and Mathematics
2018.09 - 2020.07

University of Liverpool

Bachelor of Science

Computer Science.

Algorithm and OptimisationData ScienceIntelligent Information Systems
2016.09 - 2018.07

Xi'an Jiao-Tong Liverpool University (XJTLU)

Bachelor of Science

Information and Computer Science.

Algorithm and Data StructureDiscrete Mathematics and ProbabilisticCalculus

Selected Publications

CVPR (Oral, top 0.8%)

Deep generative model based rate-distortion for image downscaling assessment

Yuanbang Liang*, Bhavesh Garg, Paul L. Rosin, Yipeng Qin

IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024

ICML (Spotlight)

Efficient Precision and Recall Metrics for Assessing Generative Models using Hubness-aware Sampling

Yuanbang Liang*, Jing Wu , Yu-Kun Lai , Yipeng Qin

The Forty-first International Conference on Machine Learning, 2024

ICML (Spotlight)

Exploring and Exploiting Hubness Priors for High-Quality GAN Latent Sampling

Yuanbang Liang*, Jing Wu , Yu-Kun Lai , Yipeng Qin

The Thirty-ninth International Conference on Machine Learning, 2022