I used matlab GA toolbox to solve an integer programming problem. The problem has some binary variables.
I used nonlinear constraints such as x*(1-x) = 0 for binary variables, but matlab outputs real values for these variables.
One another problem is that final solution is not feasible! I used this line of code:
options = gaoptimset(options,'CreationFcn', #gacreationlinearfeasible);
But matlab still generating no feasible solutions.
A friend suggested using inequality constraints instead of equality ones, but that failed.
Then there is two problems. 1) say matlab about binary variables, 2) generating feasible solutions.
How can I use matlab GA for my problem?
I'm not sure it is best solution, but I solved my problem by substitute constraints for penalty coefficients in fitness function.
After all, as a suggestion, anyone who has same problem can try GAlib (C++ genetic library) instead of matlab.
Related
My task is to model a certain physical problem and use matlab to solve it's differential equations. I made the model but it seems far more complex than what I've learned so far so I have no idea how to solve this.
The black color means it's a constant
I assume that by "solve" you seek a closed form solution of the form x(t) = ..., z(t) = ... Unforunately, it's very likely you cannot solve this system of differential equations. Only very specific canonical systems actually have a closed-form solution, and they are the most simple (few terms and dependent variables). See Wikipedia's entry for Ordinary Differential Equations, in particular the section Summary of exact solutions.
Nevertheless, the procedure for attempting to solve with Matlab's Symbolic Math Toolbox is described here.
If instead you were asking for numerical integration, then I will give you some pointers, but you must carry out the math:
Convert the second order system to a first order system by using a substitution w(t) = dx/dt, allowing you to replace the d2x/dt2 term by dw/dt. Example here.
Read the documentation for ode15i and implement your transformed model as an implicit differential equation system.
N.B. You must supply numerical values for your constants.
I have a linear program with order N^4 variables and order N^4 constraints. If I want to solve this in AMPL, I define the constraints one by one without having to bother about the exact coefficient matrices. No memory issues arises. When using the standard LP-solver in Matlab however, I need to define the matrices explicitly.
When I have variables with four subscripts, this will lead to a massively sparse matrix of dimension order N^4 x N^4. This matrix won't even fit in memory for non trivial problem sizes.
Is there a way to get around this problem using Matlab, apart from various column generation/cutting plane techniques? Since AMPL manages to solve it, I suppose they're either automating some kind of decomposition, or they somehow solve the LP without explicitly working with this sparse monster matrix.
Apart from sparse mentioned by m.s. you can also use AMPL API for MATLAB. It is especially useful if you already have an AMPL model and want to work with it from MATLAB.
Converting my comment into an answer:
MATLAB supports sparse matrices using the sparse command which allows you to build your constraint matrix without exceeding memory limits.
I am looking for a way to fit my experimental data by a theoretical model which is described by a non-linear differential equation.
Unfortunately this latter can only be solved numerically (by solving this second degree, non-linear differential equation).
I manage to solve the differential equation for a set of parameters using the ode45 Matlab solver but now I want to find the proper fit parameters of the model. Also, I may have to mention that my ode45 is initiated at z=zmax (max being large so I can assume it is infinity) by y(zmax)=y0 and yprime(zmax)=yprime0 and I solve backward (from zmax to z=0).
I am quite new to this kind of numerical problems, are there classical ways to solve such problems?
Does anyone knows if there is a Matlab procedure which would help me solve this? On which principles is it based/constructed? (if possible I'd like to know the theoretical trick to solve this problem in a smart way, not by trying all the possible sets of parameters which would be very time consuming (I have 5 fit parameters!).
Thank you for your precious help!
You have facy methods in the Optimization Toolbox. In case you don't have access to it, you could do it manually by:
Selecting a cost function between the experimental and model data. For example, mean-squared-error.
Doing heuristic optimization of the cost function. For example, Nelder-Mead method.
I have a function fun(x,y,z), such that say, x=1:10, y=50:60, z=100:105. Which optimization method (and how) I can use to get the minimum of this function, for example, where (x,y,z)=(3,52,101). I am working in Matlab.
Thank you for any help
Algorithms
There are many many algorithms out there that you can use for direct search optimization such as Nelder-Mead, Particle Swarm, Genetic Algorithm, etc.
I believe Nelder-Mead is a simplex optimization method which is used by fminsearch function in MATLAB.
Also, there is Genetic Algorithm which comes with MATLAB Global Optimization toolbox. You may want to give that a try as well.
Particle Swarm Optimization (PSO) is another direct search method that you can use. However, there is no official toolbox for Particle Swarm method built by Mathworks. The good news is there is quite a few PSO toolbox developed by other people. I personally have used this one and am quite happy with the performance. Its syntax is similar to Genetic Algorithm functions that come with Global Optimization Toolbox.
Discrete Optimization
Regarding your question that you are looking for a set of integer values namely x,y, and z corresponding to the minimum objective function value, I would add a part at the beginning of the objective function that rounds the variables to the closest integers and then feeds them to your main function fun(x,y,z). This way you would discretize your function space.
I hope my answer helps.
The following equation is to be solved for M by MATLAB:
(Atemp/At)^2=1/M^2*((2/(gamma+1))*(1+(gamma-1)*M^2/2))^((gamma+1)/(gamma-1))
It is not possible to solve this equation symbolically. In Maple it is easily possible to solve such an equation implicitly; now, is there also a pre-made function in Matlab that does this for me? I could program one myself, but as my skills are limited, its performance would not fit my needs.
I would try using fzero, or if that encounters problems because of complex values/infinities, fminbnd.