Coordinate-Descent on As it was https://galoishlee.github.io/tags/coordinate-descent/ Recent content in Coordinate-Descent on As it was Hugo zh-CN maocred@gmail.com (Halois) maocred@gmail.com (Halois) This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. Wed, 13 May 2026 08:00:00 +0800 从零开始理解坐标下降：一个参数一个参数地优化 https://galoishlee.github.io/mle-coordinate-descent-from-scratch/ Wed, 13 May 2026 08:00:00 +0800maocred@gmail.com (Halois) https://galoishlee.github.io/mle-coordinate-descent-from-scratch/ <p><strong>Reading:</strong> 抛硬币 10 次得 7 次正面——正面概率的最可能估计是 0.7。但如果模型有几百个参数呢？MLE 可能没有闭式解。<strong>坐标下降</strong>（coordinate descent）也许是解决这类问题最简单的思路：一次只优化一个参数，固定其他的，反复直到收敛。本文从抛硬币出发，经过线性回归，一直走到高维 Lasso——展示坐标下降从入门到核心应用的全景。</p>