Thanks a lot for Ushnish 's kind reply.
I am considering to apply simulated annealing algorithm for this
question.However, there may be just one issue - sensibility , that is to say,
for the totally same system of the function / constraints, the solution may be
different while runningthe program twice.
As is know, simulated annealing algorithm is based on randomly searching.
Theoretically, it will get the global optimized solution, but since in
practice, it is difficult to matchthe all theoretical conditions , then local
optimized solution will be got. Then I think different local optimized solution
may be got for different run.Is there any comment?
Best Wishes !
On Friday, November 20, 2015 2:02 AM, Ushnish De <ushn...@adatao.com> wrote:
Have you looked into Gurobi? Supposedly it can solve QCP (quadratically
constrained programming) problems which is what your example problem looks
like, based on your line 3. However if the problem is non convex then it'll be
hard to solve in most cases.
On Thu, Nov 19, 2015, 9:42 AM 'Zhiliang Zhu' via All ADATAO Team Members
<t...@adatao.com> wrote:
Hi all,
I have some optimization problem, I have googled a lot but still did not get
the exact algorithm or third-party open package to apply in it.
Its type is like this,
Objective function: f(x1, x2, ..., xn) (n >= 100, and f may be linear or
non-linear)Constraint functions:
x1 + x2 + ... + xn = 1, 1)
b1 * x1 + b2 * x2 + ... + bn * xn = b, 2)
c1 * x1 * x1 + c2 * x2 * x2 + ... + cn * xn * xn = c, 3)
<- nonlinear constraint
x1, x2, ..., xn >= 0 .
To find the solution of x which lets objective function globally or locally the
biggest.
I was thinking about to apply gradient descent or Levenberg-Marquardt algorithm
to solve it, however, the two are used for none constraint.I also considered
Lagrange multiplier method, but the system of gradient equations is nonlinear,
which seems difficult to solve,
Which algorithm would be proper to apply here, and is there any open package
like breeze for it?Any comment or link will be helpful.
Thanks a lot in advance!
Zhiliang
