target outputs in classification using neural networks

I want to use the following codes that is developed for cost sensitive neural network. I would be appreciated if you could let me know the followings:

1- Why it is necessary to use "LabelFormatConvertion" function at the first and the end?Is there any problem if our "TrainLabel" will be 1*n?

2- Why it is necessary to use "normalize" function for the output?

3-Is it necessary to use the two aforementioned function if "patternnet" would be employed?

http://lamda.nju.edu.cn/code_CSNN.ashx

Answer 1

[Greg Heath]

>I want to use the following codes that is developed for cost sensitive neural network. I would be appreciated if you could let me know the followings:

What do you mean by "cost sensitive" ?

>1- Why it is necessary to use "LabelFormatConvertion" function at the first and the end?Is there any problem if our "TrainLabel" will be 1*n?

I do not understand to what you refer.

2- Why it is necessary to use "normalize" function for the output?

Because the target is normalized? Otherwise I do not understand your question.

3-Is it necessary to use the two aforementioned function if "patternnet" would be employed?

No.

Hope this helps.

Greg