machine learning - ANN Training Set Scaling -

machine learning - ANN Training Set Scaling -

i working on ann using backpropagation @ moment, classification task. little confused normalizing info set using(i not have strong stats/probability background).

a sample of info shown below:

5.1, 3.5, 1.4, 0.2, 1 5.2, 2.7, 3.9, 1.4, 2 5.9, 3.0, 5.1, 1.8, 3

where lastly token of each class.

now, using sigmoid transfer function, network cannot output value greater 1, info needs normalized/scaled.

my first question; need scale both features , class, or class?

my sec question, there 'de-facto' or commonly used method of doing such scaling?

regards, jack hunt

it's recommended scale features. scaling should straightforward scaled_feature = (feature - min(featurearray))/(max(featurearray) - min(featurearray)).

so first attribute column, new info be: (5.1-5.1)/(5.9-5.1); (5.2-5.1)/(5.9-5.1); (5.9-5.1)/(5.9-5.1)

machine-learning neural-network backpropagation