In most classifications (e.g. logistic / linear regression), the term bias is ignored during regularization. Will we come to a better classification if we do not adjust the term bias?
Example:
Y = aX + b
Regularization is based on the idea that redefinition is not Ydue to what ais "overly specific", so to speak, which is usually manifested by large values of elements a.
Y
a
b , . , - .
b
, : Y = aX + b, a , b .