Current Research Projects

  • image

    Saddle Point Techniques and Minimax Optimization

    Saddle point techniques have been widely used in many situations as a powerful tool to alleviate the difficulties with nonsmoothness, constraints, high dimensions, and other issues associated with optimization problems. This often leads to solving some kind of minimax (saddle point) optimization that are fundamentally different from minimization problems. We would like to develop scalable and stable algorithms and complexity results of saddle point optimization problems with exploitable structure.

    Selected Publications

    [1] Niao He, Anatoli Juditsky, and Arkadi Nemirovski, “Mirror Prox Algorithm for Multi-Term Composite Minimization and Semi-Separable Problems,” Journal of Computational Optimization and Applications, 61(2), 275-319, 2015.

    [2] Niao He, and Zaid Harchaoui, “Semi-proximal Mirror-Prox for Nonsmooth Composite Minimization,” Neural Information Processing Systems (NIPS), 2015.

    [3] Niao He and Zaid Harchaoui, “Stochastic Semi-Proximal Mirror Prox,” NIPS 8th International Workshop on Optimization for Machine Learning, 2015.

    [4] Bo Dai, Niao He, Yunpeng Pan, Byron Boots, Le Song, “Learning from Conditional Distributions via Dual Kernel Embeddings,” Artificial Intelligence and Statistics (AISTATS), 2017.

    [5] Bo Dai, Albert Shaw, Lihong Li, Lin Xiao, Niao He, Jianshu Chen, Le Song, “Smoothed Dual Embedding Control”, NIPS Deep Reinforcement Learning Symposium, 2017.

  • image

    New Optimization Paradigms for Large-Scale Diffusion Models

    Point processes, have garnered a surge of interest in modeling temporal dynamics, information diffusion, and recurrent behaviors for a wide spectrum of applications in finance, social networks, healthcare, and etc. The development of efficient inferential analysis and decision-making for such models from real-time and large-scale event data, however, has far from reaching the same level of maturity as that for Gaussian models. The project aims to break the modeling and computation limitations of current theory and practice with point process models and significantly advances a variety of applications.

    Selected Publications

    [1] Nan Du, Yichen Wang, Niao He, and Le Song, “Time-sensitive Recommendation From Recurrent User Activities,” Neural Information Processing Systems (NIPS), 2015.

    [2] Niao He, Zaid Harchaoui, Yichen Wang, and Le Song, “Fast Optimization for Non-Lipschitz Poisson Likelihood Models,” arXiv, 2016.

    [3] Yingxiang Yang, Jalal Etsami, Niao He, and Negar Kiyavash, "Online Learning for Multivariate Hawkes Processes," Neural Information Processing Systems (NIPS), 2017.

  • image

    Fundamental Limits of Conditional Stochastic Optimization

    In the wake of recent breakthroughs in artificial intelligence, there has been a prominent transition of interests and demands from classical (single-stage) stochastic optimization to multi-stage stochastic programming. In contrast to classical stochastic optimization, multi-stage stochastic problems are known to suffer from the curse of dimensionality, for which efficient universal oracle-based algorithms are not readily available. The goal of this research is to build bridges from classical stochastic optimization to multi-stage stochastic problems by developing an understanding of the fundamental limits of an intermediate class of optimization problems - conditional stochastic optimization.

  • image

    CRRC: Artificial Intelligence in High Speed Railway System

    Our project “Research on Key Technology of Rail Transit-Based Wireless Sensor Intelligence Data” is funded by China Railway Rolling Stock Corporation (CRRC). Together with Professor Chenhui Shao's MechSE lab, we are working on developing deep learning and sensor fusion methodologies that will enhance the intelligence and automation of rolling stock, especially high-speed trains.