0
PINN 与 Deep Ritz：像有限差分与有限元，但关键不在网格
从同一个 Poisson 方程推导强形式残差与变分能量，逐项说明 PINN、Deep Ritz 与有限差分、有限元之间的相似与差异。
标签: 科学计算
0
把已知方程编成张量图：PolyNets给仿真团队出的算力题
一家做设备参数扫描的工程团队，要把同一套动力学模型在一万个初始条件下各算一遍。方程和参数都已经写进求解器，瓶颈是反复积分以及数据在CPU、GPU之间搬运。此时再收集轨迹、训练一个替代模型，未必划算：训练费用、长时间误差和外推失败都要算在总账上。10月14日发布的预印本《The Physics Is the Network: Fast, Training-Free Neural Emulation
0
CERN 为什么转向 Debian：大型科学设施操作系统迁移的工程账
CERN 计划把部分加速器控制系统从 CentOS Linux 迁移到 Debian。表面看，这是一项发行版替换；放在约 2200 台控制计算机和约 17000 台受控设备的规模下，它更接近一次基础设施再认证。加速器控制系统连接电源、磁体、真空、射频、束流诊断和安全相关设备，许多计算机还安装定制 PCI 或 PCIe 板卡。系统稳定不仅关系到办公效率，也关系到昂贵实验设施能否按计划运行。
0
PhysicsNeMo 技术解析：面向工程仿真的物理机器学习栈
PhysicsNeMo 提供神经算子、图网络、PINN、科学数据管线和分布式 GPU 组件，用于训练工业代理模型。本文梳理其技术栈、方法选择、数据与验证要求，并说明它与传统 CFD、FEA 和试验体系的合理分工。
0
MATLAB Agentic Toolkit到底解决了什么问题？一篇看懂MathWorks如何把工程软件接进AI Agent工作流
MathWorks新发布的MATLAB Agentic Toolkit，本质上不是一个简单的“让AI会写MATLAB代码”的插件，而是一套把MATLAB能力、工程知识和AI代理工作流绑定起来的基础设施。它一边通过MATLAB MCP Core Server给AI代理接上可执行的MATLAB环境，一边通过skills把MATLAB工程师的最佳实践注入代理，让AI不只是“会写几行.m文件”，而是能以更接近真实工程团队的方式完成建模、测试、诊断、应用开发与工具箱调用。
0
如果把科学计算教材重写一遍，Cartesian 可能就是正确方向
Cartesian 把算法教材做成了可执行、可回放、可交互的学习软件。本文拆解其技术路线，并进一步讨论：这种模式是否适合数值分析与科学计算类教科书写作？答案是，极其适合，而且可能比数据结构教材更有潜力。
0
Verification in Scientific Computing
Verification in Scientific Computing: from Pristine to Practical to Perimeter-Extending Joseph M. Powers Department of Aerospace and Mechanical Engineering University of Notre Dame ASME V&V 2020 V
0
Verification in Scientific Computing
Verification in Scientific Computing: from Pristine to Practical to Perimeter-Extending Joseph M. Powers Department of Aerospace and Mechanical Engineering University of Notre Dame ASME V&V 2020 V
0
Verification in Scientific Computing
Verification in Scientific Computing: from Pristine to Practical to Perimeter-Extending Joseph M. Powers Department of Aerospace and Mechanical Engineering University of Notre Dame ASME V&V 2020 V
0
Verification in Scientific Computing
Verification in Scientific Computing: from Pristine to Practical to Perimeter-Extending Joseph M. Powers Department of Aerospace and Mechanical Engineering University of Notre Dame ASME V&V 2020 V
0
Verification in Scientific Computing
Verification in Scientific Computing: from Pristine to Practical to Perimeter-Extending Joseph M. Powers Department of Aerospace and Mechanical Engineering University of Notre Dame ASME V&V 2020 V
0
Verification in Scientific Computing
Verification in Scientific Computing: from Pristine to Practical to Perimeter-Extending Joseph M. Powers Department of Aerospace and Mechanical Engineering University of Notre Dame ASME V&V 2020 V
0
Verification in Scientific Computing
Verification in Scientific Computing: from Pristine to Practical to Perimeter-Extending Joseph M. Powers Department of Aerospace and Mechanical Engineering University of Notre Dame ASME V&V 2020 V
0
Verification in Scientific Computing
Verification in Scientific Computing: from Pristine to Practical to Perimeter-Extending Joseph M. Powers Department of Aerospace and Mechanical Engineering University of Notre Dame ASME V&V 2020 V
0
Verification in Scientific Computing
Verification in Scientific Computing: from Pristine to Practical to Perimeter-Extending Joseph M. Powers Department of Aerospace and Mechanical Engineering University of Notre Dame ASME V&V 2020 V
0
Verification in Scientific Computing
Verification in Scientific Computing: from Pristine to Practical to Perimeter-Extending Joseph M. Powers Department of Aerospace and Mechanical Engineering University of Notre Dame ASME V&V 2020 V