Someone pointed out that my “fun linear algebra problems” post would be more useful with solutions. Here’s a solution to problems 1 and 2. (My favorite solution to problem 1 part (c) requires some group theory, so it isn’t suitable for beginning linear algebraists, but there are mildly more annoying ones using only linear algebra.)
Problem 1
Let
$\mathbb{F}$
be the finite field of$p^k$
elements. Let$V$
be a$n$
-dimensional vector space over$\mathbb{F}$
.a. How many linear transformations
$T: V \to V$
are there?b. How many of these are invertible?
(a) Linear maps $T: V \to V$
are in bijection with $n \times n$
matrices over the field $\mathbb{F}$
. Each matrix entry has $p^k$
possibilities so the total number of matrices is $$p^{kn^2}.$$
(b) Recall that $T$
is invertible iff it maps a basis to another basis. Then there are as many invertible maps as there are bases of $V$
. Imagine picking such a basis $e_1, \dots, e_n$
sequentially; we will count how many possibilities there are for each choice.
When picking vector $e_i$
, it may not be in the span of $e_1, \dots, e_{i-1}$
, or we get linear dependence again. That span is a vector space of dimension $i-1$
, so it has $|\mathbb{F}|^{i-1}$
elements. Any of the other vectors in $V$
are available to pick, so there are $|V| - |\mathbb{F}|^{i-1}$
possibilities. (This also holds for $i=1$
if we define the span of the empty set to be the zero-dimensional vector space.)
Multiplying the number of possibilities for each vector together, we get $\prod_{i=1}^n |\mathbb{F}|^n - |\mathbb{F}|^{i-1}$
or, simplifying, $$\prod_{i=0}^{n-1} (p^{kn} - p^{ki}).$$
(c) Let $GL(V)$
denote the set of invertible maps $V \to V$
; let $SL(V)$
denote the subset of such maps of determinant 1. Note that $GL(V)$
is a group with composition as the group law, and that $\det : GL(n) \to \mathbb{F}^\times$
is a group homomorphism whose kernel is $SL(V)$
. Then $SL(V)$
is a normal subgroup of $GL(V)$
of index $|\mathbb{F}^\times| = p^k - 1$
, so $$|SL(V)| = |GL(V)|/(p^k-1).$$
Problem 2
Let
$f_n$
be the$n$
th Fibonacci number.
- Find a linear transformation
$T : \mathbb{R}^2 \to \mathbb{R}^2$
such that$T(f_{n-1}, f_{n}) = (f_n, f_{n+1})$
for all$n$
. Write down the matrix of$T$
with respect to the standard basis.- Find an algorithm for determining the
$n$
th Fibonacci number in$O(\log n)$
time rather than the typical$O(n)$
(assuming addition and multiplication are$O(1)$
).- What are the eigenvalues of
$T$
?- Derive the closed-form expression for the
$n$
th Fibonacci number.
(a) The desired map is $T(x, y) = (y, x+y)$
; its matrix is $$M = \left(\begin{array}{cc}0 & 1 \\ 1 & 1\end{array}\right)$$
(b) We have $f_n$
is the first coordinate of $T^n(0, 1)$
, so we can compute $f_n$
if we can compute $M^n$
. Multiplying $2 \times 2$
matrices is constant time, so using repeated squaring we can compute $M^n$
in logarithmic time.
(c) The characteristic polynomial of $M$
is $x^2 - x - 1$
, so the eigenvalues are $\frac{1 \pm \sqrt{5}}{2} = \{\phi, 1-\phi\}$
(where $\phi$
is the golden ratio).
(d) The eigenvectors of $T$
are $(1, \phi)$
, corresponding to the eigenvalue $\phi$
, and $(1, 1-\phi)$
with eigenvalue $(1-\phi)$
. We can decompose $(0, 1) = \frac{(1, \phi) - (1, 1-\phi)}{\sqrt{5}}$
. Then $$T^n(0, 1) = T^n\left(\frac{(1, \phi) - (1, 1-\phi)}{\sqrt{5}}\right) = \frac{\phi^n(1, \phi) + (1-\phi)^n(1, 1-\phi))}{\sqrt{5}},$$
and for the closed-form expression we get $$f_n = \frac{\phi^n - (1-\phi)^n}{\sqrt{5}}.$$
Comments