Data pre-processing.

Hai all, I have a set of data taken by extracting texture features. I have to use it for classification. Each feature has 4 classes. But they do not have very good inter-class separation.

if i multiply each element of a class by a constant , and if i repeat it for every class what would be the result ? I tested it once, i found that after this process the inter-class separation was very good. But is it right to do this ?Please help.. Thanks..

[Oleg Komarov]

No idea what you're talking about: http://www.mathworks.com/matlabcentral/answers/6200-tutorial-how-to-ask-a-question-on-answers-and-get-a-fast-answer

[Jan]

I agree that the question is not clear.

According to all I know about classifications in general, the multiplication by a factor cannot increase the separation as long as the uncertainty is not limited by the resolution. This is the idea behind the microskop and the teleskop. If the objects overlap, a multiplication has no positive effect. When the objects don't overlap, the must be a process for separation, which works independent from a common factor of magnification.