Você está na página 1de 4

Kronecker Product and the vec Operator

Denition 1. Let A be an n p matrix and B an m q


matrix. The mn pq matrix
AB =

a
1,1
B a
1,2
B a
1,p
B
a
2,1
B a
2,2
B a
2,p
B
.
.
.
.
.
.
.
.
.
.
.
.
a
n,1
B a
n,2
B a
n,p
B

is called the Kronecker product of A and B. It is also called


the direct product or the tensor product.
Some properties of the Kronecker product:
1. A(B C) = (AB) C associativity,
2. A(B +C) = (AB) + (AC) distributivity
(A+B) C = (AC) + (B C),
3. For scalar a, a A = Aa = aA,
4. For scalars a and b, aAbB = ab AB,
5. For conforming matrices, (AB)(C D) = AC BD,
6. (AB)
T
= A
T
B
T
, (AB)
H
= A
H
B
H
,
7. For vectors a and b, a
T
b = ba
T
= b a
T
(note aa
T
= a a
T
),
EE 527, Kronecker Product 1
8. For partitioned matrices, [A
1
, A
2
] B = [A
1
B, A
2
B],
but A[B
1
, B
2
] = [AB
1
, AB
2
],
9. For square nonsingular matrices A and B:
(AB)
1
= A
1
B
1
,
10. For mm matrix A and n n matrix B:
|AB| = |A|
n
|B|
m
,
11. tr(AB) = tr(A)tr(B),
12. rank(AB) = rank(A) rank(B).
Denition 2. The vec operator creates a column vector from
a matrix Aby stacking the column vectors of A = [a
1
a
2
a
n
]
below one another:
vec(A) =

a
1
a
2
.
.
.
a
n

.
Theorem 1.
vec(AXB) = (B
T
A) vec(X).
Proof. Let B = [b
1
b
2
b
n
] (of size m n) and X =
EE 527, Kronecker Product 2
[x
1
x
2
x
m
]. Then, the kth column of AXB is
(AXB)
:,k
= AXb
k
= A
m

i=1
x
i
b
i,k
= [b
1,k
A b
2.k
A b
m,k
A]

x
1
x
2
.
.
.
x
m


vec(X)
= ([b
1,k
, b
2,k
, . . . , b
m,k
]

b
T
k
A])vec(X)
Stacking the columns together
vec(AXB) =

(AXB)
:,1
(AXB)
:,2
.
.
.
(AXB)
:,n

b
T
1
A
b
T
2
A
.
.
.
b
T
n
A

vec(X)
= (B
T
A) vec(X).
2
Corollary: vec(AB) = (I A) vec(B) = (B
T
I) vec(A).
EE 527, Kronecker Product 3
Some Properties of the vec Operator
tr(ABC) = vec(A
T
)
T
(I B) vec(C),
and its corollary
tr(AB) = vec(A
T
)
T
vec(B).
Also
vec(aa
T
) = a a,
tr(A
T
BCD
T
) = vec(A)
T
(DB) vec(C).
EE 527, Kronecker Product 4

Você também pode gostar