报告题目：On the Robust Heterogeneous Federated Learning Algorit…

报 告 人：陈川

时 间：2025年7月12日（周六）上午9:30-11:00

地 点：国交2号楼315报告厅

简 介：Recent advances in federated learning (FL) have enabled collaborative model training across distributed, heterogeneous clients, driving innovations in privacy-sensitive domains like healthcare and finance. However, heterogeneity in data, models, and tasks introduces critical challenges, including biased aggregation, vulnerability to adversarial attacks, and communication inefficiency. This talk introduces some recent innovative methodologies to tackle these challenges, including: (1) tensor-driven optimization to capture complex cross-client correlations, (2) prototype alignment mechanisms to enforce domain-invariant feature representations.

报告题目：Deep Mutual Density Ratio Estimation with Bregman Div…

报 告 人：​孙六全

时 间：2025年7月7日（周一）下午14:00-15:00

地 点：国交2号楼315

简 介：This talk introduces a unified approach to estimating the mutual density ratio, defined as the ratio between the joint density function and the product of the individual marginal density functions of two random vectors. It serves as a fundamental measure for quantifying the relationship between two random vectors. Our method uses Bregman divergence to construct the objective function and leverages deep neural networks to approximate the logarithm of the mutual density ratio. We establish a non-asymptotic error bound for our estimator, achieving the optimal minimax rate of convergence under a bounded support condition. Additionally, our estimator mitigates the curse of dimensionality when the distribution is supported on a lower-dimensional manifold. We extend our results to overparameterized neural networks and the case with unbounded support. Applications of our method include conditional probability density estimation, mutual information estimation, and independence testing. Simul...

报告题目：随机特征方法

报 告 人：韩东鹏

时 间：2025年7月5日（周六）下午17:00-18:00

地 点：国交2号楼315会议室

简 介：