Math 104 Fall 2019

Homework 8
Due Thursday, December 5 (before class)

Happy Thanksgiving!
Homework should be submitted via gradescope.

1. Show that if A is diagonizable, then det(A) is equal to the product of its eigenvalues.
In fact, the result above is always true whether A is diagonalizable or not. If we factor the characteristic
polynomial as
χ(λ) = (λ1 − λ)(λ2 − λ) . . . (λn − λ),
we have det(A) = λ1 λ2 . . . λn . Show why this is so.
2. Laub, Chapter 10, exercise 7.
3. Suppose you have n vectors x1 , . . . , xn in Rm . In class, we have seen that the first principal component
is the unit-normed vector w ∈ Rm so that the projections of those vectors onto w have maximum
variance.
Another way to look at this is as follows: consider a line L going through some point x0 ∈ Rm and
with some orientation u ∈ Rm , kwk = 1 (the equation of this line is x0 + tw where t is a scalar). Now
consider the line that is closest to the point in the sense that it minimizes
n
X
|distance(xi , L)|2
i=1

(the sum of squares of the distances between the xi ’s and the line).
(a) Show that the slope of the closest line is the first principal component.
Pn
(b) Show that this line goes through the average vector x̄ = i=1 xi .
4. Do either (a) or (b) (not both).
(a) Laub, Chapter 10, excercise 8.
(b) Write a function that computes the best rank-k approximation to a given matrix A. Apply your
function to the data set from problem set #5 after de-meaning the columns/questions.