本报告围绕国家在高性能矩阵计算方面的重大战略需求，针对传统存算分离架构的算力瓶颈，系统汇报了团队在基于磁存储器（MRAM）的存算一体芯片方向的研究成果。工作主要从“数模混合电路”与“随机自旋计算”两条技术路径展开。其一，提出并流片验证了一种非易失性数字式存算一体架构（nvDCIM），通过单元设计、阵列拓扑与训练算法的协同优化，实现了最高16比特精度且精度无损的矩阵-向量乘运算。实测芯片算力达4.42 TOPS，能效为11...

Deep neural networks have become the cornerstone of modern machine learning, yet their multi-layer structure, nonlinearities, and intricate optimization processes pose considerable theoretical challenges. In the first part of the talk, I will review recent advances in random matrix analysis that shed new light on these complex ML models. Starting with the foundational case of linear regression, I ...

主讲人学术背景：刘永强，中国科学技术大学几何与物理研究中心特任教授。研究领域为代数几何和代数拓扑，主要研究方向是代数簇奇异点的拓扑及Hodge理论。已在Adv. Math，Crelle和Math. Ann等国际数学杂志上发表多篇论文。 讲座题目：Kahler weighted right-angled Artin group 讲座摘要：It is a question of Serre to characterize finitely presented groups that can serve as the fundamental group of a compact Kahler manif...

随着循证医学的发展，医学研究及其日常的临床应用日趋方法化、规范化。同时，创新试验设计 （innovative trial design）、系统综述和元分析（systematic reviews and meta-analysis）越来越多的应用到前沿医学科研当中。主讲人结合其在英国从事医学科研研究的经历，着重介绍医学相关的统计分析方法和研究发展方向。 嘉宾介绍：黄超，统计学学者，英国赫尔大学准教授（Reader in Statistics）。南开大学统计学本硕，英国曼彻斯特大...

Deep neural networks have become the cornerstone of modern machine learning, yet their multi-layer structure, nonlinearities, and intricate optimization processes pose considerable theoretical challenges. In the first part of the talk, I will review recent advances in random matrix analysis that shed new light on these complex ML models. Starting with the foundational case of linear regression, I ...

In this talk, I will present our recent work on the optimality of Approximate Message Passing (AMP) algorithms for spiked matrix models with rotationally invariant noise. We introduce a new AMP algorithm that employs a matrix denoiser—acting on the eigenvalues of the observed matrix—and an iterate denoiser—applied to the AMP iterates. The resulting dynamics admit a simple state-evolution character...