π 곡λΆνλ μ§μ§μνμΉ΄λ μ²μμ΄μ§?
[μ νλμ] Least Square (μ΅μμμΉ) λ³Έλ¬Έ
[μ νλμ] Least Square (μ΅μμμΉ)
μ§μ§μνμΉ΄ 2021. 11. 25. 17:24211125 μμ±
Least Square
provides the approximiation of the solution for the over-determined system
: arg κ³μ λ°κΏμ£Όλ μΈμ (νλΌλ―Έν° 맀κ°λ³μ)
: vector Ax λ column space Col A μμ μλ€
: Col Aμμ bμ κ°μ₯ κ°κΉμ΄ ν¬μΈνΈ
normal equation
if C = A(T)A is invertible (=singular)
: A κ° square matrix μ΄λ©΄
A, A(T)A, AA(T) λͺ¨λ invertible or singular
Another Derivation of Normal Equation
symmetric matrix κ²½μ°
- νμ diagonoalizable μ
- eigen value λ real number
- eigen vector λ€μ λͺ¨λ orthogonal μ
A(T)A λ νμ symmetric matrix
: μ΅μν positive semi-definite, λ§μΌ -μΈ eigen value μμ μ positive definite
- symmetric matrix κ° positive definite κ²½μ° eigen value λ€μ λͺ¨λ μμ
=> μ¦ invertible
- symmetric matrix κ° positive semi=definite κ²½μ°, eigen value λͺ¨λ 0
'π©βπ» μ»΄ν¨ν° ꡬ쑰 > etc' μΉ΄ν κ³ λ¦¬μ λ€λ₯Έ κΈ
[λ΄κ° λ³Ό μ μ©ν λͺ¨μ] VS Code (0) | 2022.01.29 |
---|---|
[λ΄κ° λ³Ό μ μ©ν λͺ¨μ] λͺ λ Ήμ΄ (0) | 2022.01.29 |
[λ΄κ° λ³Ό μ μ©ν λͺ¨μ] μ»΄νμΌ μ΅μ (0) | 2022.01.29 |
[v0.1_κ·Έλ¦Όν νμ΄μ§ ꡬμΆνκΈ°] (0) | 2022.01.18 |
[μ νλμ] Diagonalization κ³Ό Orthogonal μ 무μμΌκΉ (0) | 2021.11.25 |