My research focuses on deep learning theory, PDE learning, numerical PDEs and image processing. Deep learning theory: I develope approximation theories and statistical learning theories of deep neural networks on various problems, especially when data have some low-dimensional structures.

Proceedings of the International Symposium on Automation and Robotics …

Hao Liu. Google DeepMind. Verified email at google.com - Homepage. Machine Learning Neural Networks Deep Learning. Articles Cited by Public access. Title. ... World model on million-length video and language with ringattention. H Liu, W Yan, M Zaharia, P Abbeel. arXiv e-prints, arXiv: 2402.08268, 2024.

Mathematical Methods in Image Processing and Inverse Problems: IPIP 2018 … On the numerical solution of nonlinear eigenvalue problems for the Monge-Ampère operator. Commun. Comput. Phys.

My recent interest is to design a graph foundation model leveraging LLMs. OneForAll enables a single GNN to work on classification tasks across various datasets and task-levels. GOFA introduces a generative graph foundation model that combines GNN and LLM layers for large-scale pretraining through graph-level next-token prediction.

Hao Liu, Xue-Cheng Tai, Roland Glowinski. An operator-splitting method for the Gaussian curvature regularization model with applications to surface smoothing and imaging. . SIAM Journal on Scientific Computing 44 (2), A935-A963, 2022 .

Hao Liu, Shingyu Leung. An Alternating Direction Explicit Method for Time-Dependent Evolution Equations with Applications to Fractional Differential Equations . Methods and Applications of Analysis, Special Issue in Honor of Roland Glowinski, 26(3), 249-268, 2019 .

We show that stochastic gradient Markov chain Monte Carlo (SG-MCMC) – a class of scalable Bayesian posterior sampling algorithms proposed recently – satisfies strong differential privacy with carefully chosen step sizes. We develop theory on the performance of the proposed differentially-private SG-MCMC method.

Dr. Hao Liu is an assistant professor in Hong Kong Baptist University. Liu received the Bachelor degree in applied and computational mathematics from Hong Kong Baptist University in 2014, and the PhD degree in applied mathematics from The Hong Kong University of Science and Technology in 2018.

Towards Predicting Urban Land Use Changes: A Dynamic Graph Alignment Perspective. ACM Transactions on Intelligent Systems and Technology, Accepted. [ICLR] Lu Dai, Yijie Xu, Jinhui Ye, Hao Liu, Hui Xiong. SePer: Measure Retrieval Utility Through The …