13th Penrose generalized inverse matrix (I)
I. Penrose generalized inverse matrix definition and Existence
In a broad sense, the original concepts or results are promoted. We know that the inverse matrix concept is intended for non-singular (or full-rank) phalanx. Therefore, this concept can be extended to: (1) singular square matrix; (2) non-square matrix. In fact, the Penrose generalized inverse matrix covers two situations.
For full-rank phalanxA,AAndAA=AA = ITherefore, of course
The four equations clearly established for full-rank phalanx constitute the inspiration of Penrose's generalized inverse.
- Penrose definition: SetAC, IfZCMake the following four equations true,
Aza =,Zaz = z,(AZ)=AZ,(Za) =Za
It is calledZIsAMoore-Penrose (generalized) Inverse of,,A.
The preceding four equations are called the Penrose equation (I), (ii), (iii), and (iv ).
- Existence and Uniqueness of the Moore-Penrose Inverse
Theorem: GivenAC,ABoth exist and are unique.
Proof: existence.AC, All have a MatrixUC,VCEnable
UAV = d= That isA = UDV
Where, yesAAAll non-zero feature values.
At this timeZ = VU
CThen
=
That is,
Where,
Uniqueness: SetZ, YIf the four Penrose equations are met
That is,ZIs unique.
This proof actually provides a construction method of Moore-Penrose inverse. We can see from the uniqueness: (1) WhenAWhen it is a full-rank square matrix, (2) the limit is actually quite strict, narrow, and more relaxed.
- {}-Inverse definition: if the first equation in the Penrose equation is satisfied, it is calledZIsA-Inverse, As, all. -Inverse common class, but real
The following five categories are commonly used:A{1 },A{1, 2 },A{1, 3 },A{1, 4 },A{1, 2, 3, 4} =
Ii. {1}-inverse nature
Theorem:
Proof: the rank of the matrix = row Rank = column rank.
(1) it must be a linear independent vector group. Therefore, other column vectors can be expressed:
VisibleABLinear Combination of all column vectors. That is
(2) likewise. A Linear Independent Vector Group. Therefore, other column vectors can be expressed:
Visible,ABEach row of vector is a linear combination, so
Together
Theorem: set, then
(1)
(2)
(3)S, TIt is a reversible matrix and correspondsAMultiplication, then
(4 )(
(5 )()
(6)
(7)
(8)
Proof: (1)
(2),... apparently true.
,
(3)
(4)
(5)
Again
Likewise,
(6 ),
Likewise
Method: Write
BothMDimension column vector, then
That is
Hence
Likewise
Method:
Again
In, replace it
(7) Take it as an example.
That isMOrder full-rank reversible phalanx, which exists.
Power Equality:, multiply by, get
(8)
That is, enable
Pair
And,
That is, make.
Theorem: MatrixAWhen and only whenAIt has a unique {1} inverse when it is a full-rank matrix.
Job: P306 3, 4, 5