0
|
1387 Views
|
2 Replies
|
2 Total Likes
View groups...
Share

# FeatureExtract function with "StandardizedVector" do not show the standard devaition of one

Posted 6 months ago
 I am wondering what kind of scaling is applied to the FeatureExtract function with "StandardizedVector" option. The output does not show the standard devaition of one for each columns. This function is used in NetEncoder["FeatureExtractor"] also. x={{1., 2., 3.}, {4., 5., 6.}, {7., 8., 9.}, {10., 11., 12.}, {13., 14.,15.}} xS01 = FeatureExtract[x, "StandardizedVector"]; xS02 = Standardize[x]; StandardDeviation /@ {xS01, xS02} Mean /@ {xS01, xS02} // Chop {{1.11803, 1.11803, 1.11803}, {1., 1., 1.}}  Thanks. Attachments:
2 Replies
Sort By:
Posted 6 months ago
 Thank you for clarification. Could you please provide few references? I am not aware of this biased estimator for the variance.
Posted 6 months ago
 It uses the biased estimator for the variance x = {{1., 2., 3.}, {4., 5., 6.}, {7., 8., 9.}, {10., 11., 12.}, {13., 14., 15.}} xS01 = FeatureExtract[x, "StandardizedVector"]; xS02 = Standardize[x]; xS03 = Sqrt[Length[x]/(Length[x] - 1)] xS02; StandardDeviation /@ {xS01, xS02, xS03} Mean /@ {xS01, xS02, xS03} // Chop (* {{1., 2., 3.}, {4., 5., 6.}, {7., 8., 9.}, {10., 11., 12.}, {13., 14., 15.}} {{1.11803, 1.11803, 1.11803}, {1., 1., 1.}, {1.11803, 1.11803, 1.11803}} {{0, 0, 0}, {0, 0, 0}, {0, 0, 0}} *) 
Community posts can be styled and formatted using the Markdown syntax.