본문 바로가기
Deep Learning lecture

ML lec 5-1: Logistic Classification의 가설 함수 정의

by xangmin 2020. 4. 28.
반응형

Classification

· Spam E-mail Detection : Spam or Ham

· Facebook feed : show or hide

· Credit Card Fraudulent Transaction detection : legitimate/fraud

 

0, 1 encoding

· Spam Detection : Spam(1) or Ham(0)

· Facebook feed : show(1) or hide(0)

· Credit Card Fraudulent Transaction detection : legitimate(0) or fraud (1)

 

Linear regression

· We know Y is 0 or 1

 -H(x) = Wx + b

· Hypothesis can give values large than 1 or less than 0

 

Sigmoid

g(z) = 1 / (1+exp(-z))

< logistic function >

 

Logistic Hypothesis

z = Wx

H(x) = 1 / (1 + exp(-W^T*X))

반응형

댓글