袁亚湘-优化方法,从瞎子爬山谈起.ppt

  1. 1、本文档共75页,可阅读全部内容。
  2. 2、有哪些信誉好的足球投注网站(book118)网站文档一经付费(服务费),不意味着购买了该文档的版权,仅供个人/单位学习、研究之用,不得用于商业用途,未经授权,严禁复制、发行、汇编、翻译或者网络传播等,侵权必究。
  3. 3、本站所有内容均由合作方或网友上传,本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺!文档内容仅供研究参考,付费前请自行鉴别。如您付费,意味着您自己接受本站规则且自行承担风险,本站不退款、不进行额外附加服务;查看《如何避免下载的几个坑》。如果您已付费下载过本站文档,您可以点击 这里二次下载
  4. 4、如文档侵犯商业秘密、侵犯著作权、侵犯人身权等,请点击“版权申诉”(推荐),也可以打举报电话:400-050-0827(电话支持时间:9:00-18:30)。
查看更多
袁亚湘-优化方法,从瞎子爬山谈起

* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * 小人物 ?? 大人物 Hotelling(1885-1973) : “But we all know the world is nonlinear.” Von Neumann(1903-1957): “Mr. Chairman, Mr Chairman, if the speaker doesn’t mind, I would like to reply for him. The speaker titled his talk `linear programming’ and carefully stated his axioms. If you have an application that satisfies the axioms, well use it. If it does not, then don’t.” 线性规划:内点法 Interior Point Method (Karmarkar, 1984) xk 0 ? 内点 N. Karmarkar(1957- ) November 19, 1984 内点法 与 罚函数 min cTx s.t. A x = b x = 0 Log-barrier function: min cTx - ∑ log (xi) s.t. A x = b KKT ? Newton’s Step 内点法和平面几何 内点法 与 SDP Semi-Definite Programming (SDP) min C, X s.t. A, X = b X = 0 C,X = Trace CTX when X=xxT ? C,X = xTCx 交替方向法 交替方向方法 min f(x) x=(x1,…, xn) 固定 xj (j≠i), 对 xi 进行优化 例子: Gauss-Seidel 方法 求解 Ax = b 交替方向方法 -- L1 优化 min ||x||1 s.t. Ax = b 罚函数方法: 交替方向方法 -- L1 优化 对于: min ||Ay-b||2 可以 用梯度法 下降; 对于 min ||x||1 |x+i | = |xi| - δ (shrinkage operator) Now this supreme wisdom, united to goodness that is no less infinite, cannot but have chosen the best. For as a lesser evil is a kind of good, even so a lesser good is a kind of evil if it stands in the way of a greater good; and the would be something to correct in the actions of God if it were possible to the better. As in mathematics, when there is no maximum nor minimum, in short nothing distinguished, everything is done equally, or when that is n

文档评论(0)

wyjy + 关注
实名认证
内容提供者

该用户很懒,什么也没介绍

1亿VIP精品文档

相关文档