N***l
发帖数: 52 | 1 i think the problem is the optimizer for f and g may be different,
and if they ARE different, any composite objective function can only
be a ``guess" of the solution, actually there is no such thing as
``the" solution in your problem.
As in the f-g case, you are maximizing the difference, but the thing
is the maximizer for the difference function may not simultaneously
maximize f and minimize g, all you know is it maximize the difference.
I think there is no best solution in general, but in your |
|
N***l
发帖数: 52 | 2 I think there are counter examples (bad examples) for both of your
composite ``optimal" objectives.
Hi there,
Sorry I can't input Chinese right now.
Please help me with the following question. Thank you so much!
To solve a problem, I have two objectives:
1. choose xi to max f(xi),
2. choose xi to min g(xi).
where f(x) and g(x) are two functions of x.
Since there are two objectives, I want to combine them to one objective
function to ease the problem.
Two possible choices of the objective functi |
|
s***t
发帖数: 113 | 3 You also need to think about how to solve the resulted model.
In general, f(x) - g(x) is much easier than f(x)/g(x). You can also
model the problem in 2-level optimization problem, e.g.,
min g(x)
s.t. max f(x)
s.t. x \in X
The resultant problem is however usually very hard to solve. |
|
m*******l
发帖数: 220 | 4 刚接触optimization不久，问个问题：一个因变量，两个自变量，自变量的边界已知，求
因变量的最小值。有哪些heuristic的算法可以使用？Matlab能做吗?
谢了。 |
|
s***t
发帖数: 113 | 5 I would suggest to try some general modelling environment first.
Metaheuristics, e.g., SA, GA, etc, requires significant insight and
extensive experiment with parameters, and may not be easy to implement.
Take a look at GAMS (www.gams.com) for some local and global optimization
solvers. For your problem size, evaluation version should suffice. Even
for larger problems, you can try NEOS online server
http://neos.mcs.anl.gov/neos/, which is also free. |
|
s***t
发帖数: 113 | 6 what type of optimization problem it is? constraint or non-constraint.
My impression is that with only 2 variables, you can probably try some
brute-force method to figure out the global optimum rather than seek
local optima.
求 |
|
a**********w
发帖数: 1 | 7 Hey, there,
I am desperate in need of help on an optimization problem using Matlab. The
idea is that I need to find the minimum for a function f(b1,b2). I have
theoretically proven that the minimum exists and local is also global.
However when I use fminsearch in Matlab using a simulation, the solution is
no where near the true values. Also, I don't know how to incorporate the
gradient & Hessian matrix.
I know that everyone is busy and your time is valuable. If it helps, I am
willing to compensa |
|
f******k
发帖数: 297 | 8 try fminbnd or fmincon. fminsearch solves the unconstrained optimization
problem. you can provide gradient and/or Hessian if you are using fmincon.
The
is |
|
w*******a
发帖数: 27 | 9 找个盗版的，
肯定有optimization tollbox
呵呵，就看你们能用盗版的不了 |
|
r****o
发帖数: 1950 | 10 看到convex optimization的一个结论是
convex set上的convex函数的local minimum即为global minimum
convex set上的concave函数的local maximum即为global maximum
请问这个global minum或global maximum是否考虑了边界点的情况。
例如f(x) concave，定义域为[a,b]，假如f(x)在(a,b)上存在local maximum，
那么这个local maximum是(a,b)上的global maximum呢，还是[a,b]上的global
maximum呢？
多谢指教。 |
|
A*******r
发帖数: 768 | 11 这样吧，
你把
Fundamentals of Convex Analysis (Grundlehren Text Editions)
Jean-Baptiste Hiriart-Urruty, Claude Lemarechal
看一遍吧
这个是精简本，实在不行就看那个两卷本的
那个
Convex Analysis
by Ralph Tyrell Rockafellar
也可以看看
这个是气宗的练法
资质好的三年有成，资质一般的五年七年吧
剑宗的练法简单一点
Convex Analysis and Optimization
by Dimitri Bertsekas
Boyd 的那本
Vandenberghe 的关于 semidefinite programming 的书
Luenberg的那本书
Chatal + Dantzig的书
Horn and Johnson的矩阵轮
Wright + Ye的内点法的书
bla bla bla
看你具体做什么哈
一年有成 |
|
X*****r
发帖数: 2521 | 12 就是objective function是time varying的
有没有那种optimization或者其他的数学model是deal这种time varying objective fu
nction的？
多谢多谢！ |
|
c****n
发帖数: 86 | 13 Do you know what it is known as (这种工作被称为什么) if I want to determine
the optimal form of a function given variables. For example,
x and y are the variables of function f (x, y).
f(x,y) can be f(x,y) = x + y or f(x,y) = x / y or the other forms.
We hope to maximize f(x,y) with certain constraints.
Thanks! |
|
A*******r
发帖数: 768 | 14 你的描述不准确
可以是functional equations
也可以是optimal control
或者是我不知道的东西
determine |
|
s****h
发帖数: 921 | 15 烦请推荐几本nonlinear optimization的书。
想弄清楚到底有哪些类的方法。 |
|
g****t
发帖数: 31659 | 16 袁亚湘写过一本，很不错看。
烦请推荐几本nonlinear optimization的书。
想弄清楚到底有哪些类的方法。 |
|
l*****a
发帖数: 119 | 17 通俗易懂的
Numerical Optimization 2nd, by Jorge Nocedal and Stephen Wright
经典的
Nonlinear Programming 2nd，by Dimitri P. Bertsekas
中文的：
袁亚湘 孙文瑜 书名忘记了 此书后来出了英文版 |
|
o*******i
发帖数: 396 | 18 【 以下文字转载自 EE 讨论区 】
发信人: orioniori (猎人), 信区: EE
标 题: Nonlinear Constrained Optimization 问题求教
发信站: BBS 未名空间站 (Sun Mar 2 14:18:31 2008)
在下是新手，不才， 有初级问题请教达人们：
目标函数 F(x1,x2,....xn)可以closeform表示出来，是个非线性函数（很多erfc函数
和指数运算，非Quadratic）。
约束函数T(x1,x2,....xn)也是一个非线性非quadratic的函数，并且可以closeform表
示出来。
现在给定Tmax上限常量，要求：
[x1,x2,....xn] = argmax{F(x1,x2,....xn)}
s.t.
T(x1,x2,....xn) <= Tmax
有什么比较好的方法，除了穷举之外？
有什么比较好 的reference或者tutorial可以参考?
十分感谢。 |
|
D*******a
发帖数: 3688 | 19 好像有个journal of optimization theory and applications，不知道如何
Methods |
|
l*****a
发帖数: 119 | 20 哪个journal review的比较快呢？
Optimization Methods and Softwares 怎么样？
那请问哪个journal比较合适呢？
如果Nemirovski Nestorov Todd这种算超一流 Yinyu Ye这样的算一流 暂且算我
老板算二流吧
Methods |
|
A*******r
发帖数: 768 | 21 我看过的文章中只有1篇来自
optimization M & S
不敢讲
一个办法是查查你老板在哪些杂志当editior-in-chief, associate or area editor
或者在board里面的， 这个review起来一般就会快一点了
送给认识的人的话相互给面子审的快点的概率大一点
要不就查查 Interior-point method 最近的文章发在哪里
现在这个多半是跟SDP有关的吧
国内好像不是很多人做SDP
其实杂志没那么重要啦
除非像 math programming那样，等文章发出来估计连博士都拿到了哈哈 |
|
X*****r
发帖数: 2521 | 22 比如变量是一个psd matrix
有专门对付这种变量是一个matrix的optimization类型吗？
多谢指点！ |
|
r******r
发帖数: 74 | 23 一个关于X的函数的optimization，有一个constraint为：X是一个半正定对称矩阵，计
为：X>=0
那么在写Lagrangian表达式的时候，关于这个条件的式子为 tr（MX）＝0，
那么这个M矩阵在KKT条件中要满足什么呢？
是不是M必须是个半正定的对称矩阵？
多谢！ |
|
m*******t
发帖数: 168 | 24 小弟目前在学习MATLAB编程解决OPTIMIZATION的问题. 比如: support vector machine
, fisher linear discriminant. 目前,没有找到一本适合学习的书.
求高手推荐一本. 谢谢! |
|
|
x******a
发帖数: 6336 | 26 Øksendal, Bernt(N-OSLO); Reikvam, Kristin(N-OSLO)
Viscosity solutions of optimal stopping problems. (English summary)
Stochastics Stochastics Rep. 62 (1998), no. 3-4, 285–301.
my email: l******[email protected]
多谢。 |
|
w******T
发帖数: 71 | 27 Journal of Industrial and Management Optimization
Pages: 847 - 860, Volume: 6 , Issue: 4 , November 2010
Two-person knapsack game
my email: h*********[email protected]
Thanks a lot. |
|
m*****n
发帖数: 93 | 28 各位版上大神，
弱弱求助关于optimization和OR的比较好的教材 文献review。我感觉在混合整数非线
性优化（MINLP）的各种解法，非凸集逼近等算法，总是看不懂啊。
~~~ |
|
m*****n
发帖数: 93 | 29 各位版上大神，
弱弱求助关于optimization和OR的比较好的教材 文献review。我感觉在混合整数非线
性优化（MINLP）的各种解法，非凸集逼近等算法，总是看不懂啊。
~~~ |
|
m*p
发帖数: 1331 | 30 【 以下文字转载自 Quant 讨论区 】
发信人: mbp (Mac Book Pro), 信区: Quant
标 题: 数学问题求教，类似 portfolio optimization.
发信站: BBS 未名空间站 (Sat Oct 27 19:12:46 2012, 美东)
这里math牛人多，希望能得到解答。
不知道有没有人知道这个问题的名字？ thanks!
Given a set of N time series, group them into K groups, such that each group
has minimal variance. |
|
m*p
发帖数: 1331 | 31 【 以下文字转载自 Quant 讨论区 】
发信人: mbp (Mac Book Pro), 信区: Quant
标 题: 数学问题求教，类似 portfolio optimization.
发信站: BBS 未名空间站 (Sat Oct 27 19:12:46 2012, 美东)
这里math牛人多，希望能得到解答。
不知道有没有人知道这个问题的名字？ thanks!
Given a set of N time series, group them into K groups, such that each group
has minimal variance. |
|
d*****1
发帖数: 1837 | 32 the model can be solved as dynamic optimization, such as
min |s1(M) - k1| + |s2(M) - k2|
\dot{s}_0 = f(x)
\dot{s}_1 = x*f(x)
\dot{s}_2 = x^2*f(x)
s_0(0) = 0
s_0(M) = 1
s_1(0) = 0
s_2(0) = 0 |
|
m******t
发帖数: 273 | 33 I am not familar with dynamic optimization.
Why ODE needs to be used here ?
Thanks |
|
o*******w
发帖数: 349 | 34 I just tidy up your question as follows, Hope it helps.
Given μ_i, σ_i, c_i, X0, how to minimize the quantity over {Xi}
∑_i^n σ_i^2 Xi^2 + ∑ (c_i - μ_i) Xi - ∑ c_i X(i-1) + c_{n+1}*X_n
=>
∑_i^n Ai (X_i - Bi)^2 - ∑ c_i {X_{i-1} - B_{i-1}} + constant_i
where Ai, Bi are new known constants.
The original problem is thus reduced to optimization of this,
∑ Ai (X_i - bi)^2 - ∑ c_i (X_{i-1} - b_{i-1})
i.e. this
∑ Ai X_i'^2 - ∑ c_i X_{i-1}' := F... 阅读全帖 |
|
r*****t
发帖数: 286 | 35 ☆─────────────────────────────────────☆
meetinnc (zerg) 于 (Mon Jan 29 12:24:26 2007) 提到:
需要写程序解一个quadratic optimization问题( quadratic objective with linear
constraints). 不知道有没有free的比如open source的package可以用? 谢谢指点?
☆─────────────────────────────────────☆
wenge (文革) 于 (Mon Jan 29 12:33:47 2007) 提到:
Google for OOQP
☆─────────────────────────────────────☆
longhei (offer offer offer) 于 (Mon Jan 29 12:39:19 2007) 提到:
OOQP is good and free
linear
☆─────────────────────────────────────☆
|
|
y*****e
发帖数: 18 | 36 what frustrated you is just why you need study, research, test, compare, etc
as a phd candidate.
as far as i know, at least half of hedge funds have no systematic portfolio
optimizations. Just based on experieces on sector, cap, country, etc.
in the other half, at least half is using commercial software, such as barra
, lehman live, api, etc.
not too many know markowitz, bayesian... |
|
p****u
发帖数: 2596 | 37 you can choose some course like "Optimization tools and software" |
|
w********n
发帖数: 2 | 38 For phd, theoretically, it's related to calculus of variation, convex
analysis, and optimal control, check whether your institution provides such
courses. For master, just google online notes to read and put more effort on
your programming skills. |
|
i*****r
发帖数: 1302 | 39 比如用mean-variance或者别的optimization, 举个例子
A: $60/share,
B: $20/share
C: $20/share
总资产$1000, 出来的weight是60%,-20%,60%
但实际中,如果在broker里面trade,我买了A 10股, short了B 10股之后我的buying
power应该只有200块了吧, 那怎么再买30股C呢?
还有那些像long 120%, short 150%之类的,实际操作中不可能的啊 |
|
b***k
发帖数: 2673 | 40 ☆─────────────────────────────────────☆
jeannie (Nokia) 于 (Sun Jun 8 23:26:00 2008) 提到:
谢谢
☆─────────────────────────────────────☆
echostate (AI) 于 (Mon Jun 9 00:42:04 2008) 提到:
Convex Optimization by Stephen Boyd and Lieven Vandenberghe
Definitely the first choice.
BTW,
You can find the free e-book, excellent slides, some exercise, and very
useful software in the author's website at http://www.stanford.edu/~boyd/cvxbook/
☆─────────────────────────────────────☆
jeannie (Nokia) 于 |
|
s**********r
发帖数: 8153 | 41 我懂optimization也懂matlab,LZ这么合一起，我蒙了~~~ |
|
I***e
发帖数: 1136 | 42 it optimizes expected log return.
payoff
as |
|
k**u
发帖数: 698 | 43 【 以下文字转载自 Trading_System 俱乐部 】
发信人: liliwater (lyrist), 信区: Trading_System
标 题: Portfolio optimization
发信站: BBS 未名空间站 (Thu Feb 11 21:17:33 2010, 美东)
method：Mean-CVaR
portfolio : 30 stocks in DJI over past 10 years
benchmark : a modified DJI
strategy : long only and long/short
rebalance : monthly
transaction fee : not considered
results:
long only:
Net Performance % to 2010-02-28:
1 mth 3 mths 6 mths 1 yr 3 yrs 5 yrs 3 yrs p.a. 5 yrs p.a.
Portfolio 0.04 -1.84 5.05 24.74 -3.98 23.8 |
|
i*****r
发帖数: 1302 | 44 Stan Uryasev的CDar和CVaR? 我两年前就做了,就是Linear programming
optimization搞到底就这么些东西,无非是改改objective function和constraint,自己
也可以发明一个什么的
这东西不能以偏概全,你把asset和benchmark全换了再backtest一遍可以就是相反结果 |
|
p****u
发帖数: 2596 | 45 除了几个大公司,一般没有单独做portfolio Optimization的组把.我觉得这种属于
quant developer 把. 利润的大头会分到做Alpha这块,portfolio construction 分
小头阿........ |
|
i*****r
发帖数: 1302 | 46 bayesian是prior,posterior这类东西么? 怎么用在optimization中? black-litterman
不就是么? 操最讨厌bayesian这个词看到就tmd烦 |
|
i*****r
发帖数: 1302 | 47 靠,这个得要我命
我是实用派非搞学术的,只要个general idea 稍微编点程序都行,optimziation得给我
明确的objective,subjective function. black litterman照着paper也能编一下,再复
杂的像这种的就算了,一眼望去全是数学公式到最后说什么也不知道有意思么
其实我的问题很简单,比如我现在有10个股票的数据,bayesian optimization,或者
beyesian whatever到底想干嘛? 能干嘛? |
|
i*****r
发帖数: 1302 | 48 很奇妙
这样的话需要增加一个主观的prior input
运用在optimization中只不过是将mu,stdev换成conditional的了.和black litterman
基本上是一回事...
太搞了 |
|
k**k
发帖数: 61 | 49 大家好，请教个问题。
已有一组simulated returns of constituent stocks (20*10000,即20个stock,共
10000组correlated returns)。想在某个portfolio target return的constraint下通
过变化stock weights来minimize portfolio expected shortfall，不知道该如何设
objective function。主要是如何将每个scenario下portfolio expected shortfall的
计算直接体现在objective function中不知道如何处理。optimization打算直接用
fmincon函数做。大家有什么建议没有？ |
|
A*****s
发帖数: 13748 | 50 CVaR是一个无法解析表示的风险度量，你写的这个逼近方法我看不懂。。。
最简单的办法是假设portfolio回报率的分布，然后解析表示，但是这样就没意思了
成熟一点的办法是对portfolio里所有asset做correlated retun simulation来逼近真
实分布
这要求你有大量的数据基础，不是一般个人可以做得到的
我看股版天天有人号称自己在CVaR，很好奇
CVaR measure对于大投行都还是个难题，不知道这些个人是怎么实现的
不管怎么样，你的这个obj根本不连续(至少一阶不连续)，有个max存在
不知道什么神奇的optimization算法能算这个，很期待解答，别告诉我用excel solver
return |
|
