Article directory
Matrix Equation Has Solution Judgment Theorem
Judgment of solutions of linear equations
-
System of linear equations A x = b A\bold{x}=\bold{b}Ax=The necessary and sufficient condition for b to have a solutionis its coefficient matrix A and augmented matrix( A , b ) (A,\bold{b})(A,b ) have the same rankR ( A ) = R ( A , b ) R(A)=R(A,\bold{b})R(A)=R(A,b),记 r = R ( A ) = R ( A , b ) r=R(A)=R(A,\bold{b}) r=R(A)=R(A,b):
- if r = nr=nr=n has a system of equations with a unique solution
- 若 r < n r<{n} r<n equations have multiple solutions
-
For non-homogeneous linear equations, it is necessary to calculate R ( A ), R ( A , b ) R(A),R(A,\bold{b})R(A),R(A,b)
-
For homogeneous linear equations only need to calculate R ( A ) R(A)R(A)
Specialization: Judgment of Solutions of Homogeneous Linear Equations
-
This is a special case of a system of linear equations that has a solution, and the theorem can be further simplified
-
A system of homogeneous linear equations A x = 0 A\bold{x}=\bold{0}Ax=The case of 0 homogeneous equations can be understood asb \bold{b}The elements in b are all 0
-
It is easy to know that A x = 0 A\bold{x}=\bold{0}Ax=0 alwaysR ( A ) = R ( A ‾ ) = r R(A)=R(\overline{A})=rR(A)=R(A)=r , so the system of homogeneous linear equationsalways has a solution;
- We only need to compute the coefficient matrix AAA 's rankR ( A ) R(A)R ( A ) to getrrr
- if r = nr=nr=n then the system of equations has a unique solution, and it isa zero solution
- 若 r < n r<n r<A system of n equations has a non-zero solution
-
Homogeneous linear equations have a solution determination theorem: homogeneous linear equations A x = 0 A\bold{x}=\bold{0}Ax=The necessary and sufficient condition for 0 to have a solution is R ( A ) ⩽ n R(A)\leqslant{n}R(A)⩽n;
- The necessary and sufficient condition for zero solution (unique solution) is R ( A ) = n R(A)=nR(A)=n
- The necessary and sufficient condition for non-zero solutions (multiple solutions) is R ( A ) < n R(A) < nR(A)<n;
Generalization: Matrix equation AX = B AX=BAX=B has a solution
- BB hereB is a matrix of constant entries (no longer an augmented matrix of coefficient matrices)
- Theorem: Matrix equation AX = B AX=BAX=The necessary and sufficient condition for B to have a solution is R ( A ) = R ( A , B ) R(A)=R(A,B)R(A)=R(A,B)
-
Note here X , BX,BX,B is not necessarily a vector, it may be a matrix with multiple rows and columns
-
Refer to Tongji Line Code v6@p76@Theorem 6
-
prove
-
A , X , BA,X,BA,X,B are m × nm\times{n}respectivelym×n, n × l n\times{l} n×l, m × l m\times{l} m×matrix of l
-
Block X and B by column:
- X X X= ( x 1 , x 2 , ⋯ x l ) (\bold{x}_1,\bold{x}_2,\cdots \bold{x}_l) (x1,x2,⋯xl),
- B B B= ( b 1 , b 2 , ⋯ b l ) (\bold{b}_1,\bold{b}_2,\cdots \bold{b}_l) (b1,b2,⋯bl)
-
Matrix equation AX = B AX=BAX=B is equivalenttolll vectorequations(system of linear equations)
-
A X = A ( x 1 , x 2 , ⋯ x l ) AX=A(\bold{x}_1,\bold{x}_2,\cdots \bold{x}_l) AX=A(x1,x2,⋯xl)= ( A x 1 , A x 2 , ⋯ A x l ) (A\bold{x}_1,A\bold{x}_2,\cdots A\bold{x}_l) (Ax1,Ax2,⋯Axl)
-
All AX = B AX=BAX=B等价于 ( A x 1 , A x 2 , ⋯ A x l ) (A\bold{x}_1,A\bold{x}_2,\cdots A\bold{x}_l) (Ax1,Ax2,⋯Axl)= ( b 1 , b 2 , ⋯ b l ) (\bold{b}_1,\bold{b}_2,\cdots \bold{b}_l) (b1,b2,⋯bl)
- 又等价于 A x i = b i ( i = 1 , 2 , ⋯ , l ) A\bold{x}_i=\bold{b}_i(i=1,2,\cdots,l) Axi=bi(i=1,2,⋯,l ) a total oflll linear equations
- What these linear equations have in common is the same coefficient matrix AAA , which means thellThe ranks of the coefficient matrices of the l linear equations and the original matrix equationsare equal, this conclusion is very important
- The position number matrix and the constant term matrix are relatively independent
-
设 R ( A ) = r R(A)=r R(A)=r , andAAThe row echelonof A isA ~ \widetilde{A}A ,则A ~ \widetilde{A}A there is rrr non-zero lines, andA ~ \widetilde{A}A After m − r mrm−r line with all zeros
-
( A , B ) (A,B) (A,B)= ( A , b 1 , b 2 , ⋯ b l ) (A,\bold{b}_1,\bold{b}_2,\cdots \bold{b}_l) (A,b1,b2,⋯bl) ∼ r \overset{r}{\sim}∼r ( A ~ , b 1 ~ , ⋯ , b l ~ ) {(\widetilde{A},\widetilde{\bold{b}_1},\cdots,\widetilde{\bold{b}_l})} (A ,b1 ,⋯,bl )
- where A ~ \widetilde{A}A Yes AAThe row echelon formmatrixof A
- And the vector b 1 ~ , ⋯ , bl ~ \widetilde{\bold{b}_1},\cdots,\widetilde{\bold{b}_l}b1 ,⋯,bl 是 b 1 , b 2 , ⋯ b l \bold{b}_1,\bold{b}_2,\cdots \bold{b}_l b1,b2,⋯bl与 A ∼ r A ~ A\overset{r}{\sim}\widetilde{A} A∼rA The result after performing the same row transformation, namely bi ~ \widetilde{\bold{b}_i}bi does not represent a row echelon matrix
-
will be equivalent to the iiThe elementary row transformation of the augmented matrix of i linear equations is a row echelon matrix:( A , bi ) (A,\bold{b}_i)(A,bi) ∼ r \overset{r}{\sim}∼r ( A ~ , b i ~ ) {(\widetilde{A},\widetilde{\bold{b}_i})} (A ,bi ), ( i = 1 , 2 , ⋯ , l ) (i=1,2,\cdots,l) (i=1,2,⋯,l)
-
A X = B AX=B AX=B有解 ⇔ \Leftrightarrow ⇔ A x i = b i {A\bold{x}_i=\bold{b}_i} Axi=bi ( i = 1 , 2 , ⋯ , l ) (i=1,2,\cdots,l) (i=1,2,⋯,l ) have a solution
- ⇔ \Leftrightarrow ⇔ R ( A , b i ) {R(A,\bold{b}_i)} R(A,bi)= R ( A ) = r R(A)=r R(A)=r, ( i = 1 , 2 , ⋯ , l ) (i=1,2,\cdots,l) (i=1,2,⋯,l)
- ⇔ \Leftrightarrow⇔ b i ~ {\widetilde{\bold{b}_i}} bi
After m − r mrm−r components (units) are all 0( i = 1 , 2 , ⋯ , l ) (i=1,2,\cdots,l)(i=1,2,⋯,l)
- Because, if after m − r mrm−There are non-zero elements in r elements, which will causeR ( A , bi ) > R ( A ) R(A,\bold{b}_i)>R(A)R(A,bi)>R(A),导致 A x i = b i {A\bold{x}_i=\bold{b}_i} Axi=biNo solution
- And its former rrThe value of r elements will not affectR ( A , bi ) {R(A,\bold{b}_i)}R(A,bi)= R ( A ) R(A) We don't care about the establishment of R ( A )
- ⇔ \Leftrightarrow ⇔ 矩阵 ( b 1 ~ , ⋯ , b l ~ ) (\widetilde{\bold{b}_1},\cdots,\widetilde{\bold{b}_l}) (b1 ,⋯,bl ) afterm − r mrm−R line is all 0;
- ⇔ \Leftrightarrow ⇔ row echelon matrixD ~ \widetilde{D}D = ( A ~ , b 1 ~ , ⋯ , b l ~ ) (\widetilde{A},\widetilde{\bold{b}_1},\cdots,\widetilde{\bold{b}_l}) (A ,b1 ,⋯,bl ) afterm − r mrm−R line is all 0
- ⇔ \Leftrightarrow⇔ R ( D ~ ) ⩽ m − ( m − r ) = r R(\widetilde{D})\leqslant{m-(m-r)=r} R(D )⩽m−(m−r)=r , and becauseD ~ \widetilde{D}D contains A ~ \widetilde{A}A , so R ( A ~ ) = r ⩽ R ( D ~ ) R(\widetilde{A})=r\leqslant{R(\widetilde{D})}R(A )=r⩽R(D )
- ⇔ \Leftrightarrow⇔ R ( D ~ ) = r R(\widetilde{D})=r R(D )=r
- ⇔ R ( A , B ) = R ( A ) \Leftrightarrow{R(A,B)=R(A)} ⇔R(A,B)=R(A)
-
Therefore, if AX = B AX=BAX=B有解,则R ( A , B ) = R ( A ) R(A,B)=R(A)R(A,B)=R(A)
inference
- Young AX = B AX=BAX=B has a solution, thenR ( B ) ⩽ R ( A , B ) = R ( A ) R(B)\leqslant{R(A,B)}=R(A)R(B)⩽R(A,B)=R ( A ) , soR ( B ) ⩽ R ( A ) R(B)\leqslant{R(A)}R(B)⩽R ( A ) , that is,the rank of the constant term matrix is less than the rank of the coefficient matrix
- against AX = B AX = BAX=Both sides of B take the transpose operation at the same time, there isXTAT = BTX^TA^T=B^TXT AT=BT , in the same wayR ( BT ) ⩽ R ( XT ) R(B^T)\leqslant R(X^T)R(BT)⩽R(XT),即 R ( B ) ⩽ R ( X ) R(B)\leqslant{R(X)} R(B)⩽R(X)
- Finally, R ( B ) ⩽ min ( R ( A ) , R ( X ) ) R(B)\leqslant{\min(R(A),R(X))}R(B)⩽min(R(A),R(X))