Regression

Regression

Warning

This post is more than a year old. Information may be outdated.

  • 과거 관찰로 현재의 연속적인 변수(Variable)를 추출하는 것. 예) 가격
  • 이것이 선형 관계라면 선형 회귀 (Linear Regression)
  • 전체 오류( $(\texttt{prediction}-\texttt{reality})^2$)를 최소화하는 함수/변수를 찾자
  • 입력 $x \in \mathbb{R}^D$ (Features, Covariates, Contexts, Predictors, etc)
  • 출력 $y \in \mathbb{R}$ (Responses, Targets, Outcomes, etc)
  • $f: \mathbb{R}^D \to \mathbb{R}$ with $f(x) = w_0 + \sum^D_{d=1} w_d x_d = w_0 + w^T x$ (superscript $^T$ stands for transpose)
  • i.e., hyperplane, parameterized by $w = [w_1 w_2 \cdots w_D]^T$ (weights, weight vector, parameter vector, etc)
  • bias $w_0$
cho.sh