Você está na página 1de 4

Appendix A: The Essentials

of Matrix Analysis

A matrix is a rectangular collection of numbers. If there are n rows and m columns, we write
the matrix as A n×m , and the numbers of the matrix are a i j , where i gives the row number
and j gives the column number. These are also called the dimensions of the matrix. We
i=n, j =m
compactly write A = (a i j )i=1, j =1 . In rectangular form

⎡ ⎤
a 11 a 12 ··· a 1m
⎢ a 21 a 22 ··· a 2m ⎥
A=⎢
⎣ ... .. .. ⎥ .
. ··· . ⎦
a n1 a n2 · · · a nm

If n = m, we say that the matrix is square. The square matrix in which there are all 1s along
the diagonal and 0s everywhere else is called the identity matrix:

⎡ ⎤
1 0 ··· 0
⎢0 1 ··· 0⎥
In := ⎢ ⎥
⎣ ... ... · · · ... ⎦
0 0 ··· 1

Here are some facts about algebra with matrices:

1. We add two matrices that have the same dimensions, A + B, by adding the respective
components A + B = (a i j + bi j ), or

⎡ ⎤
a 11 + b 11 a 12 + b 12 ··· a 1m + b 1m
⎢ a 21 + b 21 a 22 + b 22 ··· a 2m + b 2m ⎥
A+B =⎢
⎣ .. .. .. ⎥

. . ··· .
a n1 + b n1 a n2 + b n2 · · · a nm + b nm

Game Theory: An Introduction, Second Edition. E.N. Barron.



C 2013 John Wiley & Sons, Inc. Published 2013 by John Wiley & Sons, Inc.

432
Appendix A 433

2. A matrix may be multiplied by a scalar c by multiplying every element of A by c ; that


is, c A = (c a i j ) or

⎡ ⎤
c a 11 c a 12 ··· c a 1m
⎢ c a 21 c a 22 ··· c a 2m ⎥
cA =⎢
⎣ ... .. .. ⎥ .
. ··· . ⎦
c a n1 c a n2 · · · c a nm

3. We may multiply two matrices A n×m and Bm×k only if the number of columns of A is
exactly the same as the number of rows of B. You have to be careful because not only is
A · B = B · A; in general, it is not even defined if the rows and columns don’t match
up. So, if A = A n×m and B = Bm×k , then C = A · B is defined and C = C n×k , and is
given by
⎡ ⎤⎡ ⎤
a 11 a 12 ··· a 1m b 11 b 12 ··· b 1k
⎢ a 21 a 22 ··· a 2m ⎥⎢ b 21 b 22 ··· b 2k ⎥
A·B =⎢
⎣ ... .. .. ⎥ ⎢ . .. .. ⎥ .
. ··· . ⎦⎣ .. . ··· . ⎦
a n1 a n2 · · · a nm b m1 a m2 · · · a mk

We multiply each row i = 1, 2, . . . , n of A by each column j = 1, 2, . . . , k of B in


this way:


b1 j
 ⎢ b2 j ⎥
i A · B j = a i1 a i2 · · · a im · ⎢ ⎥
⎣ ... ⎦ = a i1 b 1 j + a i2 b 2 j + · · · + a im b m j = c i j .
bm j

This gives the (i, j )th element of the matrix C . The matrix C n×k = (c i j ) has elements
written compactly as

m
cij = a ir br j , i = 1, 2, . . . , n, j = 1, 2, . . . , k.
r =1

4. As special cases of multiplication that we use throughout this book


⎡ ⎤
a 11 a 12 ··· a 1m
⎢ a 21 a 22 ··· a 2m ⎥
X 1×n · A n×m = [ x 1 x2 ··· xn ] ⎢
⎣ ... .. .. ⎥
···
. . ⎦
a n1 a n2 · · · a nm

n
n
= x i a i1 x i a i2 ··· x i a im .
i=1 i=1 i=1

Each element of the result is E (X , j ), j = 1, 2, · · · , m.


434 Appendix A

5. If we have any matrix A n×m , the transpose of A is written as A T and is the m × n


matrix, which is A with the rows and columns switched:
⎡ ⎤
a 11 a 21 ··· a n1
⎢ a 12 a 22 ··· a n2 ⎥
AT = ⎢
⎣ ... .. .. ⎥ .
. ··· . ⎦
a 1m a 2m · · · a nm

If Y1×m is a row matrix, then Y T is an m × 1 column matrix, so we may multiply A n×m


by Y T on the right to get

m ⎤
⎢ a1 j y j ⎥
⎢ j =1 ⎥
⎡ ⎤⎡ ⎤ ⎢ m ⎥
a 11 a 12 ··· a 1m y1 ⎢

⎢ a2 j y j ⎥
⎢ a 21 a 22 ··· a 2m ⎥⎢ y 2 ⎥ ⎢ ⎥
T
A n×m Ym×1 =⎢
⎣ ... .. .. ⎥ ⎢ . ⎥ = ⎢ j =1

⎥.

. ··· . ⎦⎣ .. ⎦ ⎢ . ⎥
··· ⎢ . ⎥
a n1 a n2 a nm ym ⎢ m . ⎥


⎣ ⎦
anj y j
j =1

Each element of the result is E (i, Y ), i = 1, 2, . . . , n.


6. A square matrix A n×n has an inverse A −1 if there is a matrix Bn×n that satisfies A · B =
B · A = I , and then B is written as A −1 . Finding the inverse is computationally tough,
but luckily you can determine whether there is an inverse by finding the determinant
of A. The linear algebra theorem says that A −1 exists if and only if det( A) = 0. The
determinant of a 2 × 2 matrix is easy to calculate by hand:

a a 12
det(A 2×2 ) = 11 = a 11 a 22 − a 12 a 21 .
a 21 a 22

One way to calculate the determinant of a larger matrix is expansion by minors which
we illustrate for a 3 × 3 matrix:

a 11 a 12 a 13

det(A 3×3 ) = a 21 a 22 a 23
a 31 a 32 a 33

a a 23 a a 23 a a 22
= a 11 22 − a 12 21 + a 13 21 .
a 32 a 33 a 31 a 33 a 31 a 32

This reduces the calculation of the determinant of a 3 × 3 matrix to the calculation of


the determinants of three 2 × 2 matrices, which are called the minors of A. They are
obtained by crossing out the row and column of the element in the first row (other rows
may also be used). The determinant of the minor is multiplied by the element and the
sign + or − alternates starting with a + for the first element. Here is the determinant
Appendix A 435

for a 4 × 4 reduced to four 3 × 3 determinants:



a 11 a 12 a 13 a 14

a a 22 a 23 a 24
det(A 4×4 ) = 21
a 31 a 32 a 33 a 34
a
41 a 42 a 43 a 44

a 22 a 23 a 24 a 21 a 23 a 24

= a 11 a 32 a 33 a 34 − a 12 a 31 a 33 a 34
a 42 a 43 a 44 a 41 a 43 a 44

a 21 a 22 a 24 a 21 a 22 a 23

+ a 13 a 31 a 32 a 34 − a 14 a 31 a 32
a 33 .
a 41 a 42 a 44 a 41 a 42 a 43

7. A system of linear equations for the unknowns y = (y 1 , . . . , y m ) may be written in matrix


form as A n×m y = b,  1 where b = (b 1 , b 2 , . . . , b n ). This is called an inhomogeneous
system if b = 0 and a homogeneous system if b = 0. If A is a square matrix and is
 
invertible, then y = A −1 b is the one and only solution. In particular, if b = 0, then
y = 0 is the only solution.
8. A matrix A n×m has associated with it a number called the rank of A, which is the largest
square submatrix of A that has a nonzero determinant. So, if A is a 4 × 4 invertible
matrix, then rank(A) = 4. Another way to calculate the rank of A is to row-reduce the
matrix to row-reduced echelon form. The number of nonzero rows is rank(A).
9. The rank of a matrix is intimately connected to the solution of equations A n×m y m×1 =
b n×1 . This system will have a unique solution if and only if rank(A) = m. If rank(A) <
m, then Ay m×1 = b n×1 has an infinite number of solutions. For the homogeneous
system Ay m×1 = 0n×1 , this has a unique solution, namely, y m×1 = 0m×1 , if and only if
rank(A) = m. In any other case, Ay m×1 = 0n×1 has an infinite number of solutions.

1
To simplify notation we will drop the superscript T for the transpose of a single row matrix.

Você também pode gostar