Variation of the initial matrix during fminsearch computing

[MatthieuM]

Hello, I'm using the "fminsearch" function on Scilab.
My function "f" to minimize depends of two parameters, say "p1" and "p2"

My problem is that the variation of "p1" and "p2" during the fminsearch is very low.

For example, if p1=10 and p2=50, the first try of fminsearch will be p1'=10.5 and p2'=50, which has no influence on the value of "f".

So my question is :How can i force fminsearch to increase the variation of p1 and p2 ?

Thank you very much for your help,

Matthieu