‘Can anyone tell me how I can solve this problem?’
Not definitively.
I assume you are referring to the ga function in the Global Optimization Toolbox, not a function from the File Exchange.
Much depends on the problem and the number of parameters. I usually choose a random matrix with a large (about 100) number of rows for the 'InitialPopulationMatrix', and if I know the relative magnitudes of the particular parameters, use element-wise multiplication by an appropriate vector to scale the columns of the matrix to the appropriate magnitudes. Larger matrices (more rows) are usually better at the outset, however if ga takes too long, I generally re-start it with fewer rows in the matrix. It generally requires a bit of experimentation, however if the problem is reasonably well-posed, and the initial population matrix of reasonable row length, ga usually converges relatively quickly.
