Invertible matrices¶
A square matrix \(A\) is called invertible if there exists another square matrix \(B\) of same size such that
The matrix \(B\) is called the inverse of \(A\) and is denoted as \(A^{-1}\).
Assume \(A\) is invertible, then there exists a matrix \(B\) such that
Assume that columns of \(A\) are linearly dependent. Then there exists \(u \neq 0\) such that
a contradiction. Hence columns of \(A\) are linearly independent.
Assume \(A\) is invertible, then there exists a matrix \(B\) such that
Now let \(x \in \FF^n\) be any arbitrary vector. We need to show that there exists \(\alpha \in \FF^n\) such that
But
Thus if we choose \(\alpha = Bx\), then
Thus columns of \(A\) span \(\FF^n\).
Assume \(A\) is invertible, then there exists a matrix \(B\) such that
Applying transpose on both sides we get
Thus \(B^T\) is inverse of \(A^T\) and \(A^T\) is invertible.
Assume \(A\) is invertible, then there exists a matrix \(B\) such that
Applying conjugate transpose on both sides we get
Thus \(B^H\) is inverse of \(A^H\) and \(A^H\) is invertible.
We note that
Similarly
Thus \(B^{-1}A^{-1}\) is the inverse of \(AB\).
We verify the properties of a group
- [Closure] If \(A\) and \(B\) are invertible then \(AB\) is invertible. Hence the set is closed.
- [Associativity] Matrix multiplication is associative.
- [Identity element] \(I\) is invertible and \(AI = IA = A\) for all invertible matrices.
- [Inverse element] If \(A\) is invertible then \(A^{-1}\) is also invertible.
Thus the set of invertible matrices is indeed a group under matrix multiplication.
An \(n \times n\) matrix \(A\) is invertible if and only if it is full rank i.e.
Similar matrices¶
An \(n \times n\) matrix \(B\) is similar to an \(n \times n\) matrix \(A\) if there exists an \(n \times n\) non-singular matrix \(C\) such that
Thus there exists a matrix \(D = C^{-1}\) such that
Thus \(A\) is similar to \(B\).
Let \(B\) be similar to \(A\). Thus their exists an invertible matrix \(C\) such that
Since \(C\) is invertible hence we have \(\Rank (C) = \Rank(C^{-1}) = n\). Now using here \(\Rank (AC) = \Rank (A)\) and using here we have \(\Rank(C^{-1} (AC) ) = \Rank (AC) = \Rank(A)\). Thus
Gram matrices¶
Gram matrix of columns of \(A\) is given by
Gram matrix of rows of \(A\) is given by
This is also known as the frame operator of \(A\).
Following results apply equally well for the real case.
Let \(A\) be an \(m\times n\) matrix and \(G = A^H A\) be the Gram matrix of its columns.
If columns of \(A\) are linearly dependent, then there exists a vector \(u \neq 0\) such that
Thus
Hence the columns of \(G\) are also dependent and \(G\) is not invertible.
Conversely let us assume that \(G\) is not invertible, thus columns of \(G\) are dependent and there exists a vector \(v \neq 0\) such that
Now
From previous equation, we have
Since \(v \neq 0\) hence columns of \(A\) are also linearly dependent.
Columns of \(A\) can be dependent only if its Gram matrix is not invertible. Thus if the Gram matrix is invertible, then the columns of \(A\) are linearly independent.
The Gram matrix is not invertible only if columns of \(A\) are linearly dependent. Thus if columns of \(A\) are linearly independent then the Gram matrix is invertible.
The null space of \(A\) and its Gram matrix \(A^HA\) coincide. i.e.
Let \(u \in \NullSpace(A)\). Then
Thus
Now let \(u \in \NullSpace(A^H A)\). Then
Thus we have
Rows of \(A\) are linearly dependent, if and only if columns of \(A^H\) are linearly dependent. There exists a vector \(v \neq 0\) s.t.
Thus
Since \(v \neq 0\) hence \(G\) is not invertible.
Converse: assuming that \(G\) is not invertible, there exists a vector \(u \neq 0\) s.t.
Now
Since \(u \neq 0\) hence columns of \(A^H\) and consequently rows of \(A\) are linearly dependent.
Pseudo inverses¶
Let \(A\) be an \(m \times n\) matrix. An \(n\times m\) matrix \(A^{\dag}\) is called its Moore-Penrose pseudo-inverse if it satisfies all of the following criteria:
- \(A A^{\dag} A = A\).
- \(A^{\dag} A A^{\dag} = A^{\dag}\).
- \(\left(A A^{\dag} \right)^H = A A^{\dag}\) i.e. \(A A^{\dag}\) is Hermitian.
- \((A^{\dag} A)^H = A^{\dag} A\) i.e. \(A^{\dag} A\) is Hermitian.
We omit the proof for this. The pseudo-inverse can actually be obtained by the singular value decomposition of \(A\). This is shown here.
Let \(D = \Diag(d_1, d_2, \dots, d_n)\) be an \(n \times n\) diagonal matrix. Then its Moore-Penrose pseudo-inverse is \(D^{\dag} = \Diag(c_1, c_2, \dots, c_n)\) where
We note that \(D^{\dag} D = D D^{\dag} = F = \Diag(f_1, f_2, \dots f_n)\) where
We now verify the requirements listed here.
\(D^{\dag} D = D D^{\dag} = F\) is a diagonal hence Hermitian matrix.
Let \(D = \Diag(d_1, d_2, \dots, d_p)\) be an \(m \times n\) rectangular diagonal matrix where \(p = \min(m, n)\). Then its Moore-Penrose pseudo-inverse is an \(n \times m\) rectangular diagonal matrix \(D^{\dag} = \Diag(c_1, c_2, \dots, c_p)\) where
\(F = D^{\dag} D = \Diag(f_1, f_2, \dots f_n)\) is an \(n \times n\) matrix where
\(G = D D^{\dag} = \Diag(g_1, g_2, \dots g_n)\) is an \(m \times m\) matrix where
We now verify the requirements listed here.
\(F = D^{\dag} D\) and \(G = D D^{\dag}\) are both diagonal hence Hermitian matrices.
If \(A\) is full column rank then its Moore-Penrose pseudo-inverse is given by
It is a left inverse of \(A\).
By here \(A^H A\) is invertible.
First of all we verify that it is a left inverse.
We now verify all the properties.
Hermitian properties:
If \(A\) is full row rank then its Moore-Penrose pseudo-inverse is given by
It is a right inverse of \(A\).
By here \(A A^H\) is invertible.
First of all we verify that it is a right inverse.
We now verify all the properties.
Hermitian properties: