max matrix size for linear optimization

I am using Matlab linear optimization function linprog to optimize matrix of the following dimension:

1. Objective function: 1 x 5110

2. equality constraint size: Aeq: 2347 x 5110 Beq: 2347 x 1

3. inequality constraint size: A: 567 x 5110 B: 567 x 1

4. Lower bound size: lb: 5110 x 1

5. upper bound size: ub: 5110 x 1

6. total number of matrix elements: 14,908,784

7. no of variables: 5110

all data has a double data type.

I received an error message saying OUT OF MEMORY when the program tries to form the equality matrix. I am using 32 bit machine with 32 bit MATLAB version. I was using a 2GB RAM but later changed to 4GB and got no change. so:

1. is the limitation on my hardware or on the MATLAB forming the matrices? what shall i do to solve it?

2. I also can't run my matlab program on a 64 bit machine with 64 bit matlab program. It has a problem of connecting to a database buit on Ms Access. It only reads the first three lines of the database. I figured out such problems happen when the Ms Access has an activation problem but mine is activated and i even installed Ms Acess 2013 activate it and nothing changes.

please send ur advice if u have any solution on both the above problems.

Thanks

[Ashenafi Alemu]

No answers so far. I am waiting to get some.

Answer 1

[Roger Wohlwend]

You try to solve an optimizing problem with over 5000 variables? I am not surprised that Matlab cannot do that. I think you have to re-consider your problem entirely. It does not make sense to set up an optimization with over 5000 variables. I doubt that there is a solution to such a huge optimization. Are you sure you are doing the right thing? Because in my opinion the problem is not Matlab and the hardware limitation but your model.

[Ashenafi Alemu]

Thanx for the response. With due respect either u donno MATLAB optimization well or u haven't understood my question. I have already solved a problem with more than 4000 variables but i can't pinpoint its problem regarding connection with Access (specially the 64 bit) and limitation of using the available memory by MATLAB(if any). SO i don't buy your idea.

Answer 2

[Alan Weiss]

Is your matrix Aeq sparse? I mean, does it have a large number of zeros? If so, use sparse matrices Aeq and A.

Alan Weiss

MATLAB mathematical toolbox documentation