Artificial Neural Network's input NORMALIZATION

Question

mahdi bazarghan 2011-7-12

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/11397-artificial-neural-network-s-input-normalization

评论： primrose khaleed 2014-6-18

hi every one,

I have a question regarding the normalization of kind of data I have. I have two sets of data to the ANN, training data and testing data. Both of them are having 5 data points at each input.

Training data looks like:

4.04174435161 0.00572348 -0.0153630110827 0.602838009364 -2.36559733245

3.03056109866 0.0662695 -0.0533391771878 0.55923481044 -1.91116449019

2.52071323312 0.121233 0.260306969065 0.467349509343 -1.70735726533

Testing set looks like:

2.33842672041 -0.137656 -0.0597465187944 0.448352030836 -1.00431677993

3.96620615784 0.0438668 0.219515113659 0.596424503954 -2.0851352013

1.99407739363 0.100512 0.360384341206 0.385154801184 -1.18594026984

I would like to know the way to normalize these inputs before feeding to the ANN.

Cheers Mahdi

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Hoda abuzied 2011-9-25

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/11397-artificial-neural-network-s-input-normalization#answer_22490

hi, mahdi you can normalize any data before feeding them to the network by using mapminmax command...it normalizes data so that they would fall in the range of -1 & 1(defaults values)...to do this follow the steps below:

>> i_p_data=[4.04174435161 0.00572348 -0.0153630110827 0.602838009364 -2.36559733245;

3.03056109866 0.0662695 -0.0533391771878 0.55923481044 -1.91116449019;

2.52071323312 0.121233 0.260306969065 0.467349509343 -1.70735726533]; % i just put your data in a vector and gave it a name for simplicity.

>> [nomalized_i_p,ps1]=mapminmax(i_p_data);

>> test_data=[2.33842672041 -0.137656 -0.0597465187944 0.448352030836 -1.00431677993;

3.96620615784 0.0438668 0.219515113659 0.596424503954 -2.0851352013;

1.99407739363 0.100512 0.360384341206 0.385154801184 -1.18594026984];

>>[normalized_test,ps2]=mapminmax(test_data);

>> however, if you want to set them back to their original values i.e. denormalize them :

>> training_data=mapminmax('reverse',normalized_i_p,'ps1');

>> test_data1=mapminmax('reverse',normalized_test,'ps2');

you might also, want to check this for further details

http://www.mathworks.com/help/nnet/ref/mapminmax.html

i hope this was helpful.

yours,

hoda

2 个评论
显示无隐藏无

Walter Roberson 2011-9-25

The documentation suggests that 'ps1' and 'ps2' should not be quoted in the reverse mapping call ?

primrose khaleed 2014-6-18

hi ...i used this method to normlization the input and testing data in neural network...but when apply this method the testing data dont change..why??? plz help me

请先登录，再进行评论。