Data preprocessing in ANNs

Hi, Please, could I get on helps to write a correct code' to scaling and normalizing the input data with interval [0.1-0.9] that are used as input data to neural network in matlab. regards

[Hassanean]

Dear Shashank thanks in advance for your respond. This scaling function can deal with vector matrix at one raw, in case of I have 6 raws in input matrix,please, how can solve that?

Answer 1

[Shashank]

In general, you can get a new variable in the range [a,b] using:

                   x_new = (b-a)*((x-min_x)/(max_x-min_x))+a;

Here min_x and max_x represent the minimum and maximum values of the original data and x is the value you are trying to convert.

For example, Let's say I have an array i=[1 2 3 4] which I want to scale and normalize to an interval [0.1 0.9], i can do so by using the following code snippet

i=[1 2 3 4];
i_min=min(i);
i_max=max(i);
a=0.1;                 %for range [0.1 0.9]
b=0.9;
i_new = (b-a)*((i-i_min)./(i_max - i_min)) + a

i_new=
  0.1000    0.3667    0.6333    0.9000