MIT学习笔记-Linear Algebra(一)

Teacher：Gilbert Strang

The text of the course: Introduction to Linear Algebra

Website:web.mit.edu/18.06,which got a lot of exercise from the past MATLAB codes, the syllabus for the course

class1 and class2

方程组的几何解释

画了两个图，分别表示col和row形式。row就是用正常的方程组写，然后解代表直线（二维）的交点。而col是vector相加的形式，即linear combination（这表明矩阵的col代表vector）。

elimination消元法

Can I solve Ax=b for every b?如果有解的话，就可以用消元法解出来。同样的问法 Do the linear combination of the columns fill three dimensional space?这意味着方程组的解b代表着线性空间的点。
matrix multiply 方法1、[2 5;1 3]【1 2】=1【1 2】+2【1 3】。
Ax is comb of columns of A数据量大的时候这么理解更方便。
Determinants行列式
step1:subtract 3 row1 from row2、step2:subtract 2 row2 from row3 **
[1 0 0 [1 0 0 [1 2 1
-3 1 0 * 3 8 2 = 0 2 -2
0 0 1] 0 4 1] 0 4 1]
E₂₁(E为Elementary Matrics)
a important fact move the parentheses arbitrary 任意交换括号
Problem1：image Permutation Matrics想象[a b;c d]的置换矩阵（行置换）
Problem2：image Matrix’s [1 0 0;-3 1 0;0 0 1] Inverses

Lecture 3 and Lecture 4

four way of matrix multiplication 1st:normal way, row*col
2ed: cols way; 3st: rows way; 4th:col * row(m by 1 times 1 by p equals m by p); 5th:Block way.

Inverses

A^-1（if this is exist） * A = I = A * A^-1 左逆与右逆没有区别
Gauss-Jordan：[A I]->[I A^-1]将【A I】同乘A^-1即可理解

A=LU(L、U represent lower matrix and upper matrix)

与EA=U相比，更简洁.例：E=【1 0 0；-2 1 0；10 -5 1】则L=【1 0 0；2 1 0；0 5 1】
If no row exchanges the multipliers go directly into L(应该有负号？)
**How many operations on n*n matrix A?**应以Matlab算法复杂度的角度思考。count 1/3n³ on A，count n² on b（back substitution）。

Permutation:identity matrix with reordered rows

(A^-1)^T=(A^T)^-1
number of permutation 3 by 3 matrix have 6P(Pemutation matrices)'s;n by n matrix have n! P’s
P^-1=P^T;PP^T=I
R * R^T is always symmetric

Lecture 5

Vectors spaces and subspaces

must satisfied multiply and addition

R² = all 2-dim real vectors
R³ = all columns with 3 real components

subspaces of R²
1、all of R²
2、any line through [0 0]
3、zero vector only

A = [1 3;2 3;4 1]; for columns of A, in R³ , all their combinations from subspace called column space, C(A).

Lecture 6

two ways to get the subspaces Ax=0. A=[1 2 3 4;1 1 1 1;2 3 4 5]^T
1: Columnspace: Does Ax=b have a solution for evering b? no. Which b’s allow this system to be solved? Can solve Ax=b exactly when b is in C(A)
2:Nullspace; Ax=0; N(A)=[c c -c]; check that solutions to Ax=0give a subspace always, If Av=0 and Aw=0 then A(v+w)=0 ,then A(2v)=0.
Ax=[1234],the solutions isn’t a subspace(substitution the [000] to x.).

Lecture 7

A=
1 2 2 2
2 4 6 8
3 6 8 10
echelon U=
1 2 2 2
0 0 2 4
0 0 0 0
the zero representes the second column is dependent on the eariler columns
there are two pivot columns, one free column
rank of A= # of pivots =2
# of free variables=n-r=4-2=2
special solution: two ways
1: back substitution
x₁+2x₂+2x₃+2x₄=0
2x₃+4x₄=0
X=c[-2;1;0;0]+d[2;0;-2;1]
2:R=(in MATLAB, use rref(A) to get R)
rref为reduced row echelon form
[1 2 2 2
0 0 1 2
0 0 0 0]
write:X=a*[unknow1 1 unknow2 0]^T+b*[unknow3 0 unknow4 1]
[I F
0 0] form:
[1 0 2 -2
0 1 0 2
0 0 0 0]
X=a*[-2 1 0 0]+c*[2 0 -2 1]^T
proof:
[I F] [x_point;x_free]=0
x_point=-F*x_free

Lecture 8

Complete solution of Ax=b
A=
1 2 2 2 b1
2 4 6 8 b2
3 6 8 10 b3
echelon U=
1 2 2 2 b1
0 0 2 4 b2-2*b1
0 0 0 0 b3-b2-b1
Solivability condition on b: Ax=b solvable when b in C(A).
b3-b2-b1=0 If a comb of rows of A gives zero row, then same comb. of entries of b must give 0.
To find complete sol’s to Ax=b,x=x_particular+x_nullspace
you can set all free variables to zero to get x_particular.
m by n matrix A of rank r(know r<=m, r<=n)
4 situations:
1:r=m=n
A=
1 2
2 4 ;
R=I;
1 solution
2:r=n<m
A=
1 2
2 4
3 6;
R=[I; 0];
0 or 1 solution
3:r=m<n
R=[I F];
infinite solution
4:r<m, r<n
R=
[I F
0 0 ]
0 or infinite solutions

Lecture 9

Suppose A is m by n. with m<n. Then there are nonzero solutions to Ax=0.(more unknow than equations)
Reason: There will be free variables!!!
Independence: vectorsX₁,X₂,…,X_n are independent if no combination gives zero vector (except the zero comb,all c_j=0)
c₁x₁+c₂x₂+…+c_nx_n=0
rank=n; N(A)={0};no free variables;
rank<n; yes free variables
span: Vectors V₁,…V_l span a space means: The space consists of all combs of those vectors
Basis: Basis for a space is a sequence of vectors V₁,V₂,…,V_d with 2 propertities
1、They are independent
2: They span the space
Dimension: Every basis for the space has the same number of vectorrs, the number is DEFINITION OF　DIMENSION
rank= # of pivot columns=dimension of C(A)
dim N(A)= # of free variables= n-r

Lecture 10

4 subspaces

1、columnspace C(A) in R^m
dim(C(A))=r;
2、nullspace N(A) in Rⁿ
dim(N(A))=n-r=free variables=# of special solutions
3、row space = all combs. of rows = all combs of columns of C(A^T)=C(A^T)
dim(C(A^T))=r
4、null space of A^T = N(A^T)=left nullspace of A
dim(N(A^T))=m-r
A=
1 2 3 1
1 1 2 1
1 2 3 1
R=
1 0 1 1
0 1 1 0
0 0 0 0
C( R)!=C(A): different col space || same row space
N(A^T) A^Ty=0 y^TA=0
how to solve N(A^T): E[A_m*n I_m*m]–>[R_m*n E_m*n]
EA=R
E=
-1 2 0
-1 1 0
-1 0 0 0
the yellow part is y^T. In chap 2, R was I, The E was A^-1

new vector space

All 3+3 matrices
subspaces of M || upper triangulars || symmetric matrices || diagonal matrices ( dim of this subspace is 3.)

Lecture 11

In R⁴, v=[v1;v2;v3;v4],
S= all v in R⁴ with v1+v2+v3+v4=0
=nullspace of A =[1 1 1 1]
rank=1; dim(N)=n-r=3

small world graph

Graph={nodes, edges}
(这节存的稿子没保存，淦)

Lecture 12

Incidence matrics
node1 2 3 4
A = [-1 1 0 0 edge
0 -1 1 0
-1 0 1 0
-1 0 0 1
0 0 -1 1]
row1、2、3 consist a loop
Ax=0(x=x1,x2,x3,x4) --e=Ax–>x2-x1,etc,…Potential
differences --y=Ce–>current y1,y2,y3,y4,y5 --A^Ty=f–> Kirchoff’s CL
A^TCAx=f
Tree=no loops
A^Ty=0 dimN(A^T)=2=m-r=2

Lecture 13

review
N(CD)=N(D); if C is invertible