Can we use strings as target matrix in neural network fitting app?

I am trying to build a network which has numerical inputs of radiance values, and I am trying to map them to different cloud types (which are ofcourse strings). The neural network fitting app doesnt allow me to use cell array when choosing the target matrix. So I just gave each string a different integer for now. But as expected, I am getting large MSE and the linear fit is not that great either. The maximum gradient it reaches is 0.6.

Is there a way around it somhow? Any help is much appreciated.

Here is the problem

Input matrix: several 1000s of samaples of 11 inputs Target matrix: Cloudtypes for that number of samples

Answer 1

[Walter Roberson]

If you were able to use strings, the result would not be any better, because at best it would match strings for equality. If you have a relationship between target strings, a fuzzy membership for example, then you would need to find a way of expressing that, because certainly the code would not know how to find it out from the hypothetical target strings.

Answer 2

[Greg Heath]

If you have c cloud types make the targets {0,1} c-dimensional unit column vectors from the unii matrix eye(c).

Hope this helps.

Thank you for formally accepting my answer

Greg

[Kashyapa Narenathreyas]

Sorry, I could not understand. Do you mean to make the targets logical by refering to {0,1}?

Thanks