Exercise 4.3.3

Answers

Note, multiplying P in front of a matrix permutates its rows, while multiplying P on the back of a matrix permutates its columns.

Let the size of A be m × n, and the size of B be M × N.

Consider the (i 1)M + kth row in A B, this row comes from the product of the ith row from A, ai,, with the kth row from B, i.e. bk,. This row looks like

ai1bk1,ai1bk2,,ai1bkN,ai2bk1,ai2bk2,,ai2bkN,,ai(n1)bk1,ai(n1)bk2,,ai(n1)bkN,ainbk1,ainbk2,,ainbkN

To have the same elements of bkqaip = aipbkq in the matrix of B A, they are on the row of (k 1)m + i. This row looks like

bk1ai1,bk1ai2,,bk1ain,bk2ai1,bk2ai2,,bk2ain,,bk(N1)ai1,bk(N1)ai2,,bk(N1)ain,bkNai1,bkNai2,,bkNain

So we should switch every (i 1)M + kth row in A B with the (k 1)m + ith row.

We also need switch the columns in B A as well, consider the entry of bkqaip of B A, its column position is (q 1)n + p. The element from A B with the same value, i.e. aipbkq is at the position of (p 1)N + q So we need swtich all columns between (q 1)n + p and (p 1)N + q.

Combine the two conditions together, we see that P should switch rows satisfying following condition:

(i 1)M + k = (q 1)n + p and (k 1)m + i = (p 1)N + q

  • For A B, its eigenvalues are always the products of eigenvalues from A and B, this is also true for B A, so their eigenvalues are the same.
User profile picture
2020-03-20 00:00
Comments