Home

Yueming Lyu

E-mail: yueminglyu [at] gmail.com and Lyu_Yueming [at] ihpc.a-star.edu.sg
[Google Scholar]

Brief Biography

statistical machine learning

optimization

black-box optimization

kernel methods

approximation theory

QMC theory is a fundamental part of approximation theory. The goal of QMC methods is to design good points set for integral approximation. It can be used in Bayesian Inference and kernel approximation, etc. QMC based feature maps are promising direction of random feature methods , which can reduce the time and space complexity of kernel methods (e.g., Gaussian Process and SVM). Interestingly, there is a close relationship between random feature maps and neural networks. Besides, QMC methods also have applications in sampling techniques in RL and generative models.

Black-box optimization is a subarea of optimization. It handles the cases that only function quire can be accessed. The potential applications include reinforcement learning (RL), engineering design and black-box attack, etc. The goal is to design efficient and theoretical sounded black-box optimization algorithms to improve query efficiency. These algorithms can be used for model-free RL. Moreover, the methodology of black-box optimization is also helpful for designing efficient RL methods.

Both robust learning and weakly-supervised learning are subareas of machine learning. Robust learning and weakly-supervised learning aims at addressing a more practical situation in real-life. The input collection may contain noise; the annotation may be corrupted or even expensive to acquire. It is challenging to develop smart algorithms to handle these cases.

Yueming Lyu and Ivor W. Tsang. Neural Optimization Kernel: Towards Robust Deep Learning. Preprint.
Yueming Lyu, Yuan Yuan and Ivor W. Tsang. Subgroup-based Rank-1 Lattice Quasi-Monte Carlo. Conference on Neural Information Processing Systems (NeurIPS), 2020
Yueming Lyu. Spherical structured feature maps for kernel approximation. In International Conference on Machine Learning (ICML), 2017

Yueming Lyu, Kim Yong Tan, Yew-Soon Ong, Ivor W. Tsang. Covariance-Adaptive Sequential Black-box Optimization for Diffusion Targeted Generation. Preprint
Feiyang Ye^*, Yueming Lyu^* , Xuehao Wang, Yu Zhang, Ivor W. Tsang. Adaptive Stochastic Gradient Algorithm for Black-box Multi-Objective Learning. Accepted In International Conference on Learning Representations (ICLR), 2024
Yueming Lyu. Fast Rank-1 Lattice Targeted Sampling for Black-box Optimization. In Conference on Neural Information Processing Systems (NeurIPS), 2023
Jing Li, Yuangang Pan, Yueming Lyu, Yinghua Yao, Yulei Sui, Ivor W. Tsang Earning Extra Performance from Restrictive Feedbacks. IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI)
Yueming Lyu and Ivor W. Tsang. Black-box Optimizer with Implicit Natural Gradient. In European Conference on Machine Learning. (ECML), 2021
Yueming Lyu, Yuan Yuan and Ivor W. Tsang. Efficient Batch Black-box Optimization with Deterministic Regret Bounds. Preprint.
Xingrui Yu, Yueming Lyu and Ivor W. Tsang. Intrinsic Reward Driven Imitation Learning via Generative Model. In International Conference on Machine Learning (ICML), 2020

Feng Hong, Yueming Lyu, Jiangchao Yao, Ya Zhang, Ivor Tsang, Yanfeng Wang. Diversified Batch Selection for Training Acceleration. Accepted In International Conference on Machine Learning (ICML), 2024
Feng Hong, Jiangchao Yao, Yueming Lyu, Zhihan Zhou, Ivor Tsang, Ya Zhang, Yanfeng Wang. On Harmonizing Implicit Subpopulations. Accepted In International Conference on Learning Representations (ICLR), 2024
Yueming Lyu and Ivor W. Tsang. Curriculum Loss: Robust Learning and Generalization against Label Corruption. In International Conference on Learning Representations (ICLR), 2020
Yuan Yuan, Yueming Lyu, Xi Shen, Ivor W. Tsang, Dit-Yan Yeung. Marginalized average attentional network for weakly-supervised learning. In International Conference on Learning Representations (ICLR), 2019