# I want to set a range of variable to three or four places of decimals

4 views (last 30 days)
WooYong Lee on 25 Nov 2020
Commented: Walter Roberson on 26 Nov 2020
How can i set a range of variable to three or four places of decimals? I'm going to write a function of the variable and float a graph. I want to find optimal value under specific function with the variable.

John D'Errico on 26 Nov 2020
Are you asking to create a variable that only carries 3 or 4 digits of precision? You can't choose to set an arbitrary precision, unless you use a tool like my HPF, but that would take far more effort on your part to learn than it would help you, and it has limits.
At best, you can use single precision, giving you roughly 8 digits. And Cleve Moler has posted some tools for shorter prescision, but those tools will also be limited in which functions can use them.
Or, are you asking how to plot a function over a domain that varies over 3 or 4 orders of magnitude for a plot?
That is far easier, since you can use tools like logspace, or semilogx, etc. And of course, you can always work using logs.
Walter Roberson on 26 Nov 2020
logspace(A, B, N) is the same as 10.^(linspace(A,B,N))
For example,
format long g
logspace(-2, 5, 3)
ans = 1×3
0.01 31.6227766016838 100000
10.^linspace(-2,5,3)
ans = 1×3
0.01 31.6227766016838 100000
You can do things like:
mat2str(10.^(-2:5))
ans = '[0.01 0.1 1 10 100 1000 10000 100000]'
but that does not restrict the range of the variable for optimization routines.

Walter Roberson on 26 Nov 2020
I want to find optimal value under specific function with the variable.
The MATLAB minimizers such as fmincon() and ga() all accept an options parameter, and the options parameter permits you to set a function tolerance -- so that you could stop searching when the function is changing less than a certain value, instead of comtinuing on to try to find the absolute best position.