kb/data/en.wikipedia.org/wiki/Generalized_inverse-0.md

12 KiB
Raw Blame History

title chunk source category tags date_saved instance
Generalized inverse 1/3 https://en.wikipedia.org/wiki/Generalized_inverse reference science, encyclopedia 2026-05-05T07:23:55.090630+00:00 kb-cron

In mathematics, and in particular, algebra, a generalized inverse (or, g-inverse) of an element x is an element y that has some properties of an inverse element but not necessarily all of them. The purpose of constructing a generalized inverse of a matrix is to obtain a matrix that can serve as an inverse in some sense for a wider class of matrices than invertible matrices. Generalized inverses can be defined in any mathematical structure that involves associative multiplication, that is, in a semigroup. This article describes generalized inverses of a matrix

    A
  

{\displaystyle A}

. A matrix

      A
      
        
          g
        
      
    
    ∈
    
      
        R
      
      
        n
        ×
        m
      
    
  

{\displaystyle A^{\mathrm {g} }\in \mathbb {R} ^{n\times m}}

is a generalized inverse of a matrix

    A
    ∈
    
      
        R
      
      
        m
        ×
        n
      
    
  

{\displaystyle A\in \mathbb {R} ^{m\times n}}

if

    A
    
      A
      
        
          g
        
      
    
    A
    =
    A
    .
  

{\displaystyle AA^{\mathrm {g} }A=A.}

A generalized inverse exists for an arbitrary matrix, and when a matrix has a regular inverse, this inverse is its unique generalized inverse.

== Motivation == Consider the linear system

    A
    x
    =
    y
  

{\displaystyle Ax=y}

where

    A
  

{\displaystyle A}

is an

    m
    ×
    n
  

{\displaystyle m\times n}

matrix and

    y
    ∈
    
      
        C
      
    
    (
    A
    )
    ,
  

{\displaystyle y\in {\mathcal {C}}(A),}

the column space of

    A
  

{\displaystyle A}

. If

    m
    =
    n
  

{\displaystyle m=n}

and

    A
  

{\displaystyle A}

is nonsingular then

    x
    =
    
      A
      
        
        1
      
    
    y
  

{\displaystyle x=A^{-1}y}

will be the solution of the system. Note that, if

    A
  

{\displaystyle A}

is nonsingular, then

    A
    
      A
      
        
        1
      
    
    A
    =
    A
    .
  

{\displaystyle AA^{-1}A=A.}

Now suppose

    A
  

{\displaystyle A}

is rectangular (

    m
    ≠
    n
  

{\displaystyle m\neq n}

), or square and singular. Then we need a right candidate

    G
  

{\displaystyle G}

of order

    n
    ×
    m
  

{\displaystyle n\times m}

such that for all

    y
    ∈
    
      
        C
      
    
    (
    A
    )
    ,
  

{\displaystyle y\in {\mathcal {C}}(A),}




  
    A
    G
    y
    =
    y
    .
  

{\displaystyle AGy=y.}

That is,

    x
    =
    G
    y
  

{\displaystyle x=Gy}

is a solution of the linear system

    A
    x
    =
    y
  

{\displaystyle Ax=y}

. Equivalently, we need a matrix

    G
  

{\displaystyle G}

of order

    n
    ×
    m
  

{\displaystyle n\times m}

such that

    A
    G
    A
    =
    A
    .
  

{\displaystyle AGA=A.}

Hence we can define the generalized inverse as follows: Given an

    m
    ×
    n
  

{\displaystyle m\times n}

matrix

    A
  

{\displaystyle A}

, an

    n
    ×
    m
  

{\displaystyle n\times m}

matrix

    G
  

{\displaystyle G}

is said to be a generalized inverse of

    A
  

{\displaystyle A}

if

    A
    G
    A
    =
    A
    .
  

{\displaystyle AGA=A.}

The matrix

      A
      
        
        1
      
    
  

{\displaystyle A^{-1}}

has been termed a regular inverse of

    A
  

{\displaystyle A}

by some authors.

The problem is how to choose an

    x
  

{\displaystyle x}

as the output of

    G
  

{\displaystyle G}

for every

    y
  

{\displaystyle y}

when the map

    x
    ↦
    y
    =
    A
    x
  

{\displaystyle x\mapsto y=Ax}

is not bijective.

If

    A
  

{\displaystyle A}

is not surjective, then not all

    y
  

{\displaystyle y}

's in its codomain have corresponding

    x
  

{\displaystyle x}

's via

    A
  

{\displaystyle A}

. To circumvent it, we just let

    G
  

{\displaystyle G}

map those

    y
  

{\displaystyle y}

's to arbitrary values. For example, decompose the codomain of

    A
  

{\displaystyle A}

as the direct sum of the column space

        C
      
    
    (
    A
    )
  

{\displaystyle {\mathcal {C}}(A)}

and a complement subspace, and construct

    G
  

{\displaystyle G}

as follows. For

    y
  

{\displaystyle y}

's in the former subspace, let

    G
  

{\displaystyle G}

map back to the corresponding

    x
  

{\displaystyle x}

's. For

    y
  

{\displaystyle y}

's in the latter subspace, let

    G
  

{\displaystyle G}

map them all to zero (as there're no corresponding

    x
  

{\displaystyle x}

's). For other

    y
  

{\displaystyle y}

's, decompose them as the sum of the above two components, apply

    G
  

{\displaystyle G}

respectively, then take the sum. If

    A
  

{\displaystyle A}

is not injective, then some

    y
  

{\displaystyle y}

's correspond to multiple

    x
  

{\displaystyle x}

's via

    A
  

{\displaystyle A}

. To circumvent it, we let

    G
  

{\displaystyle G}

map every

    y
  

{\displaystyle y}

to one of the

    x
  

{\displaystyle x}

's according an algorithm. For example, decompose the domain of

    A
  

{\displaystyle A}

as the direct sum of

    ker
    
    A
  

{\displaystyle \ker A}

and a complement subspace. For every possible

    y
  

{\displaystyle y}

, its preimage must be parallel to

    ker
    
    A
  

{\displaystyle \ker A}

and intersect the chosen complement subspace at a single point. Let

    G
  

{\displaystyle G}

map the

    y
  

{\displaystyle y}

to this point. If

    A
  

{\displaystyle A}

is neither surjective nor injective, we combine the above two tricks. The picture on the right is an example.

== Types == Important types of generalized inverse include:

One-sided inverse (right inverse or left inverse) Right inverse: If the matrix

    A
  

{\displaystyle A}

has dimensions

    m
    ×
    n
  

{\displaystyle m\times n}

and

        rank
      
    
    (
    A
    )
    =
    m
  

{\displaystyle {\textrm {rank}}(A)=m}

, then there exists an

    n
    ×
    m
  

{\displaystyle n\times m}

matrix

      A
      
        
          R
        
      
      
        
        1
      
    
  

{\displaystyle A_{\mathrm {R} }^{-1}}

called the right inverse of

    A
  

{\displaystyle A}

such that

    A
    
      A
      
        
          R
        
      
      
        
        1
      
    
    =
    
      I
      
        m
      
    
  

{\displaystyle AA_{\mathrm {R} }^{-1}=I_{m}}

, where

      I
      
        m
      
    
  

{\displaystyle I_{m}}

is the

    m
    ×
    m
  

{\displaystyle m\times m}

identity matrix. Left inverse: If the matrix

    A
  

{\displaystyle A}

has dimensions

    m
    ×
    n
  

{\displaystyle m\times n}

and

        rank
      
    
    (
    A
    )
    =
    n
  

{\displaystyle {\textrm {rank}}(A)=n}

, then there exists an

    n
    ×
    m
  

{\displaystyle n\times m}

matrix

      A
      
        
          L
        
      
      
        
        1
      
    
  

{\displaystyle A_{\mathrm {L} }^{-1}}

called the left inverse of

    A
  

{\displaystyle A}

such that

      A
      
        
          L
        
      
      
        
        1
      
    
    A
    =
    
      I
      
        n
      
    
  

{\displaystyle A_{\mathrm {L} }^{-1}A=I_{n}}

, where

      I
      
        n
      
    
  

{\displaystyle I_{n}}

is the

    n
    ×
    n
  

{\displaystyle n\times n}

identity matrix. BottDuffin inverse Drazin inverse MoorePenrose inverse Some generalized inverses are defined and classified based on the Penrose conditions:

    A
    
      A
      
        
          g
        
      
    
    A
    =
    A
  

{\displaystyle AA^{\mathrm {g} }A=A}




  
    
      A
      
        
          g
        
      
    
    A
    
      A
      
        
          g
        
      
    
    =
    
      A
      
        
          g
        
      
    
  

{\displaystyle A^{\mathrm {g} }AA^{\mathrm {g} }=A^{\mathrm {g} }}