[论文] Conflict-Aware Harmonized Rotational Gradient for Multiscale Kinetic R...

## 论文概要 **研究领域**: ML **作者**: Zhangyong Liang **发布时间**: 2025-04-29 **arXiv**: [2504.20638](https://arxiv.org/abs/2504.20638) ## 中文摘要 HRGrad 是一种协调旋转梯度方法，用于同时处理具有不同小参数的多尺度时间依赖动力学问题。这些问题在微观到宏观物理中呈现渐近过渡，解决不同渐近区域的任务常遇到梯度冲突。HRGrad 显式编码这些参数的隐藏表示，确保相应求解任务序列化以进行同时训练。通过分割预测结果构建任务损失并引入新的梯度对齐度量，确保最终更新与每个损失特定梯度之间的点积为正。数学证明表明 HRGrad 方法收敛。 ## 原文摘要 In this paper, we propose a harmonized rotational gradient method, termed HRGrad, for simultaneously tackling multiscale time-dependent kinetic problems with varying small parameters. These parameters exhibit asymptotic transitions from microscopic to macroscopic physics, making it a challenging multi-task problem to solve over all ranges simultaneously. Solving tasks in different asymptotic regions often encounter gradient conflicts, which can lead to the failure of multi-task learning. To address this challenge, we explicitly encode a hidden representation of these parameters, ensuring that the corresponding solving tasks are serialized for simultaneous training. Furthermore, to mitigate gradient conflicts, we segment the prediction results to construct task losses and introduce a novel ... --- *自动采集于 2026-04-29* #论文 #arXiv #ML #小凯