Solving Equations
선형방정식의 해를 체계적으로 분석하는 방법과 이를 바탕으로 한 vector space, eigenvalue, Newton's method, inverse function과 implicit function의 핵심 개념을 정리합니다.
2.1 Row Reduction
1) Linear System
선형방정식계는 matrix equation으로 표현할 수 있다.
여기서
- : coefficient matrix
- : vector of unknowns
- : vector of constants
2) Augmented Matrix
선형방정식계는 augmented matrix로 나타낼 수 있다.
Meaning: 방정식의 coefficient와 constant를 하나의 matrix로 묶어 계산한다.
3) Row Operations
해를 바꾸지 않는 기본 row operation은 세 가지이다.
- 한 row에 nonzero scalar를 곱한다.
- 한 row의 scalar multiple을 다른 row에 더한다.
- 두 row를 교환한다.
Row operation을 적용해도
의 solution set은 변하지 않는다.
4) Echelon Form
책에서 사용하는 Echelon Form은 다음 조건을 만족한다.
- 각 nonzero row의 첫 nonzero entry는 이다.
- 아래 row의 pivotal 은 위 row의 pivotal 보다 오른쪽에 있다.
- pivotal 이 있는 column의 다른 entry는 모두 이다.
- zero row는 모두 아래에 위치한다.
Meaning: 방정식의 해 구조를 바로 읽을 수 있도록 matrix를 정리한 형태이다.
5) Row Reduction
모든 matrix는 row operations를 통해 echelon form으로 변환할 수 있다.
이 과정을 Row Reduction 또는 Gaussian Elimination이라고 한다.
책에서 중요한 점은 주어진 matrix의 echelon form이 unique하다는 것이다.
2.2 Solving Linear Equations
1) Solution Classification
를 echelon form
으로 row reduce한다.
해는 세 가지 경우로 나뉜다.
2) No Solution
column에 pivotal 이 있으면
는 solution을 가지지 않는다.
Meaning: 서로 모순되는 equation이 존재한다.
3) Unique Solution
에 pivotal 이 없고 의 모든 column에 pivotal 이 있으면 unique solution을 가진다.
4) Infinitely Many Solutions
에 pivotal 이 없고 에 nonpivotal column이 있으면 infinitely many solutions을 가진다.
5) Pivotal Variable
Pivotal Variable: echelon form에서 pivotal 을 포함하는 column에 대응하는 variable.
6) Nonpivotal Variable
Nonpivotal Variable: pivotal 이 없는 column에 대응하는 variable.
Nonpivotal variables의 값을 자유롭게 선택하면 pivotal variables의 값이 결정된다.
Meaning: nonpivotal variable은 free variable이고 pivotal variable은 그것에 의존한다.
7) Invertibility Criterion
Square matrix 에 대해
따라서
가 모든 에 대해 unique solution을 가지는 것과 가 invertible인 것은 같은 조건이다.
2.3 Matrix Inverses and Elementary Matrices
1) Solving with an Inverse
가 invertible이면
의 unique solution은
이다.
2) Computing an Inverse
의 inverse는 augmented matrix
를 row reduce하여 계산할 수 있다.
만약
이면
반대로 부분이 로 row reduce되지 않으면 은 존재하지 않는다.
3) Elementary Matrix
Elementary Matrix: identity matrix에 하나의 row operation을 적용하여 얻은 matrix.
Matrix 에 elementary matrix 를 왼쪽에서 곱하는 것은 에 해당 row operation을 적용하는 것과 같다.
Elementary matrix는 항상 invertible하며 그 inverse도 elementary matrix이다.
2.4 Linear Combinations, Span, and Linear Independence
1) Linear Combination
vectors
의 Linear Combination은
형태의 vector이다.
2) Linear Independence
vectors 가 linearly independent라는 것은
의 유일한 solution이
이라는 뜻이다.
Meaning: 어떤 vector도 나머지 vector들의 linear combination으로 만들어지지 않는다.
Linear independence는 solution의 uniqueness와 연결된다.
3) Span
Span: 주어진 vectors의 모든 linear combinations의 집합.
Meaning: 주어진 vectors를 이용해 만들 수 있는 모든 vector의 집합이다.
Span은 solution의 existence와 연결된다.
4) Matrix Criterion
라고 하자.
가 linearly independent인 것은 row-reduced 의 모든 column에 pivotal 이 있는 것과 동치이다.
또한 가 을 span하는 것은 row-reduced 의 모든 row에 pivotal 이 있는 것과 동치이다.
5) Basis
Basis: vector space 를 span하면서 linearly independent인 ordered set of vectors.
Meaning: 공간의 모든 vector를 중복 없이 표현하기 위한 기본 좌표축이다.
basis를 이용하면 모든 를 unique하게
로 표현할 수 있다.
6) Dimension
Dimension: vector space의 basis에 포함된 vector의 개수.
모든 basis는 동일한 수의 vectors를 가진다.
특히
7) Orthonormal Basis
Orthonormal Basis: basis vectors가 서로 orthogonal하고 모두 unit length인 basis.
orthonormal basis에서는
8) Orthogonal Matrix
Orthogonal Matrix: columns가 orthonormal basis를 이루는 square matrix.
따라서
Orthogonal matrix는 dot product와 length를 보존한다.
2.5 Kernels, Images, and the Dimension Formula
1) Kernel
Linear transformation
의 Kernel은
Meaning: transformation에 의해 zero로 mapping되는 모든 vectors의 집합이다.
2) Image
Image:
Meaning: transformation이 실제로 만들어 낼 수 있는 모든 output의 집합이다.
Kernel과 image는 각각 domain과 codomain의 vector subspace이다.
3) Kernel and Uniqueness
이면
는 모든 에 대해 solution을 최대 하나만 가진다.
즉,
4) Image and Existence
이면 모든 에 대해
의 solution이 존재한다.
즉,
5) Basis for the Image
Matrix 의 original pivotal columns은
의 basis를 이룬다.
주의할 점은 row-reduced matrix의 columns가 아니라 original matrix의 pivotal columns를 사용한다는 것이다.
6) Dimension Formula
Linear transformation
에 대해
이다.
Meaning: domain의 dimension은 kernel과 image 사이에 나누어진다.
7) Rank
Rank: image의 dimension.
따라서 dimension formula는
으로 쓸 수 있다.
8) Existence from Uniqueness
특히
인 경우
이면
따라서 square linear system에서는
2.6 Abstract Vector Spaces
1) Vector Space
Vector Space: vectors를 더하고 scalar를 곱할 수 있으며 이러한 operations가 vector space axioms를 만족하는 집합.
대표적인 예:
그리고 function spaces도 vector space가 될 수 있다.
Meaning: vector는 반드시 숫자의 column일 필요가 없다.
2) Linear Transformation
Vector spaces 사이의 mapping
가 linear transformation이라는 것은
를 만족한다는 뜻이다.
3) Matrix with Respect to Bases
의 basis를
의 basis를
이라고 하자.
각
를 의 basis로 표현한 coefficients가 matrix 의 번째 column이 된다.
Meaning: abstract linear transformation도 basis를 선택하면 matrix로 표현할 수 있다.
4) Change of Basis Matrix
두 bases
사이의 change of basis matrix를
라고 하면
이다.
Meaning: 같은 vector를 다른 coordinate system으로 변환한다.
5) Change of Basis Formula
Linear transformation 의 matrix representation은 basis에 따라 달라진다.
특히 domain과 codomain이 같은 vector space이고 같은 basis를 바꾸는 경우
이 형태는 eigenbasis에서 매우 중요하다.
2.7 Eigenvectors and Eigenvalues
1) Eigenvector and Eigenvalue
Linear transformation
에 대해 nonzero vector 가
를 만족하면 를 Eigenvector, 를 Eigenvalue라고 한다.
Meaning: transformation을 적용해도 방향은 변하지 않고 크기만 배 되는 vector이다.
2) Eigenspace
Eigenvalue 에 대응하는 eigenspace는
Eigenvalue의 multiplicity는 책에서 이 eigenspace의 dimension으로 정의한다.
3) Eigenbasis
Eigenbasis: 모든 basis vector가 eigenvector인 basis.
Eigenbasis를 사용하면 linear transformation이 매우 단순해진다.
4) Diagonalization
의 columns가 의 eigenbasis이고 corresponding eigenvalues가
이면
따라서
Meaning: 적절한 basis를 선택하면 transformation이 각 coordinate를 단순히 scaling하는 형태가 된다.
5) Powers of a Matrix
Diagonalization이 가능하면
그리고
Meaning: matrix의 높은 power를 계산하는 문제가 scalar powers 계산으로 분리된다.
6) Independent Eigenvectors
서로 다른 eigenvalues에 대응하는 eigenvectors는 linearly independent이다.
따라서 -dimensional vector space에서는 서로 다른 eigenvalue가 최대 개 존재한다.
7) Important Limitation
모든 matrix가 eigenbasis를 가지는 것은 아니다.
Meaning: 모든 square matrix가 diagonalizable한 것은 아니다.
2.8 Newton's Method
1) Nonlinear Equation
Nonlinear system
을 직접 풀기 어려울 때 현재 guess 근처에서 를 linearize한다.
2) Newton's Method
현재 approximation을 이라고 하면
따라서
이다.
Meaning: nonlinear equation을 매 iteration마다 linear equation으로 바꾸어 푼다.
3) Practical Computation
실제 계산에서는
을 직접 계산하기보다
을 row reduction으로 풀고
으로 계산한다.
4) Lipschitz Condition
Newton's method의 convergence를 보장하려면 derivative가 너무 빠르게 변하지 않아야 한다.
여기서 은 Lipschitz ratio이다.
5) Kantorovich's Theorem
Kantorovich's theorem은 적절한 initial guess와 Lipschitz condition 아래에서 Newton sequence가 실제 root로 convergence한다는 것을 보장한다.
책의 sufficient condition 중 핵심 형태는
Meaning: initial residual, inverse derivative의 크기, derivative의 변화율이 충분히 작으면 Newton's method의 convergence를 보장할 수 있다.
2.9 Superconvergence
1) Newton's Method Convergence
Kantorovich condition에서
처럼 strict inequality가 성립하면 Newton's method는 훨씬 빠르게 convergence할 수 있다.
책에서는 이를 Superconvergence라고 부른다.
2) Main Idea
successive increment를
라고 하면 superconvergence에서는 이 값이 iteration마다 매우 빠르게 감소한다.
Meaning: Newton's method가 root에 충분히 가까워지면 정확한 digits의 수가 매우 빠르게 증가한다.
2.10 Inverse and Implicit Function Theorems
1) Local Inverse
Function
이 global하게 invertible하지 않더라도 특정 point 주변에서는 inverse를 가질 수 있다.
이를 locally invertible이라고 한다.
2) Inverse Function Theorem
가 continuously differentiable이고
가 invertible이면 는 근처에서 locally invertible하다.
즉, 근처에 inverse function
가 존재한다.
Meaning: derivative가 invertible이면 nonlinear function도 충분히 작은 영역에서는 invertible linear transformation처럼 행동한다.
3) Derivative of the Inverse
특히
4) Local vs Global Invertibility
가 모든 point에서 invertible이라고 해서 가 global하게 invertible한 것은 아니다.
Inverse function theorem은 local theorem이다.
5) Implicit Function
Equation
이 어떤 variables를 다른 variables의 function으로 결정할 때 이를 implicit function이라고 한다.
예를 들어 locally
의 형태로 나타낼 수 있다.
6) Implicit Function Theorem
이고
이라고 하자.
만약
가 onto라면 적절한 개의 pivotal variables를 나머지 개의 nonpivotal variables의 function으로 locally 표현할 수 있다.
즉,
Meaning: nonlinear equation에서도 linear equation의 pivotal / nonpivotal variable 구조가 local하게 유지된다.
7) Derivative of an Implicit Function
variables를
로 나누고
가 invertible이면
이다.
Essential Study Checklist
반드시 이해해야 할 내용:
- Linear system
- Augmented Matrix
- Row Operations
- Echelon Form
- Pivotal / Nonpivotal Variables
- Unique / None / Infinite Solutions
- Matrix Inverse
- Linear Combination
- Linear Independence
- Span
- Basis
- Dimension
- Orthonormal Basis
- Orthogonal Matrix
- Kernel
- Image
- Rank
- Dimension Formula
- Change of Basis
- Eigenvector / Eigenvalue
- Eigenbasis
- Diagonalization
- Newton's Method
- Lipschitz Condition
- Kantorovich's Theorem의 의미
- Superconvergence의 의미
- Inverse Function Theorem
- Implicit Function Theorem
- Implicit Function Derivative