Feature standardization considered harmful

Many statistical learning algorithms perform better when the covariates are on similar scales. For example, it is common practice to standardize the features used by an artificial neural network so that the gradient of its objective function doesn’t depend on the physical units in which the features are described. ... The post Feature standardization considered harmful first appeared on R-bloggers.
http://dlvr.it/S1ctV9

تعليقات

المشاركات الشائعة من هذه المدونة

How to add R {magrittr}’s %>% Pipe Operator in VSCode as Keyboard Shortcut