4 papers indexed
Ye Tian, Mengchu Li, Marco Avella Medina
Integrating information across related tasks can improve estimation and prediction in transfer, multi-task, and federated learning, but contamination and heterogeneity make robust borrowing challenging. We study a contaminated multi-task empirical risk minimization (ERM) framewor…
Xiangli Shi, Xiaomeng Zhu, Ye Tian, Yuchun Guo, Ziyang Sun, Lujie Yin, et al.
Embodied task planning asks an agent to turn a natural-language instruction into an executable sequence of actions in a physical scene, and is a building block for household, assistive, and service robots. Recent prompting-based and reinforcement-learning planners generate fluent…
Transfer learning leverages knowledge from related source domains to improve learning in a target domain. Recent theoretical advances cover a broad range of regression settings within (generalized) linear models. Despite their diversity, these methods share two common constraints…
Quanling Zhao, Jiaying Yang, Ye Tian, Josh Victoria, Zhijun Wang, Pietro Mercati, et al.
Unsupervised tabular anomaly detection requires methods that are accurate, robust across heterogeneous datasets, and computationally efficient. Classical statistical detectors are often efficient, but they usually rely on a fixed data view and a single notion of abnormality. Deep…