Q1: 20 points Q2: 15 points Q3: 35 points Q4: 30+10 points Total: 100+10 points
Question text is in black, solution text in blue
If either Cardie dominates Duncan or Duncan dominates Cardie, then Ace both does and does not dominate Biscuit.
¬X(a, c) → (X(c, d) ⊕ X(d, a))
If Biscuit does not dominate Duncan, then Cardie dominates Ace, and if Cardie does not dominate Ace, then Biscuit does not dominate Duncan.
∃u: ∀v: (u ≠ v) → (X(u, v) ∨ X(v, u))
It was important to get the "other than it" for full credit.
Note that "(u ≠ v) ∧ ..." is wrong, as "u ≠ v" is not
true for all pairs.
For any two dogs, either the first does not dominate the second or the
second does not dominate the first.
Note that this statement allows u and v to be the same dog,
which will be useful later.
∀u: ∃v: ∃w: (u ≠ v) ∧ (u ≠ w) ∧ (v
≠ w) ∧ ¬X(u, v) ∧ ¬ X(v, u) ∧ X(w, u) ∧
X(w, v)
Note that "u ≠ v ≠ w" does not state that "u ≠ w".
Here
we can use ∧ to include the not-equal conditions as we are saying
that v and w exist that satisfy the not-equal conditions
and make the rest true as well.
∃u: ∀v: (u ≠ v) → (¬X(u, v) ∧ ¬X(v, u))
First look at Statement 1, which says that if either X(c, d) or X(d,
c) is true, then X(a, b) is both true and false. So X(c, d) ∨
X(d, c) leads to a contradiction, meaning that X(c, d) and X(d, c)
are both false.
Next, Statement III is two implications, which we can combine by
Hypothetical Syllogism to get ¬X(c, a) → X(c, a), which is
only possible if X(c, a) is true.
Specializing Statement V to c and a, and using X(c, a), we get
¬X(a, c). Turning to Statement II, we know that the premise of
this implication is true and so its conclusion is true. Since X(c,
d) is false, X(d, a) must be true.
In Question 3a we prove ∀u: ¬X(u, u) from Statement V,
telling us that X(a, a), X(b, b), X(c, c), and X(d, d) are all
false.
Specializing Statement V to d and a, we find that X(a, d) is
false.
Statement VII says that one of the dogs neither dominates nor is
dominated by any other dog. Since X(c, a) and X(d, a) are true,
this dog cannot be Ace, Cardie, or Duncan and thus must be Biscuit.
Thus X(b, a), X(b, c), X(b, d), X(a, b), X(c, b), and X(d, b) are
all false. We have now established all sixteen truth values --
the
only two that are true are X(c, a) and X(d, a).
Let u be an arbitrary dog. Specializing Statement V to make both dogs u, we get ¬X(u, u) ∨ ¬X(u, u), which implies ¬X(u, u). Since u was arbitrary, we have proved ∀u: ¬X(u, u).
This was the most disappointing set of answers on the test. Most of
you never addressed the transitivity of Y, instead asserting the
transitivity of X with no justification (except some sort of
inference
from the meaning of the English word "dominate"). Here is the
correct answer:
We want to prove ∀u: ∀ v: ∀ w: (Y(u, v)
∧ Y(v, w)) → Y(u, w). So we let u, v, and w be arbitrary
dogs and assume both Y(u, v) and Y(v, w). By the definition of Y,
there
exist two dominance chains, one with first element u and last
element v, and one with first element v and last element w. We can
create a single dominance chain by letting ui,...,
uk be the first chain (so that u1 = u and
uk = v, then letting uk+1 be the second
element of the other chain, uk+2 be the third element,
and so on, including all the elements of the second chain until the
last one, which is w. This new chain establishes that Y(u, w) is
true.
It is valid because X(ui, ui+1) for every i in
the new chain is given by the conditions for one or the other of
the original chains.
For the second question, Y is not guaranteed to be
antisymmetric even if X is. (Almost no one got this right, as most
of you didn't recognize any difference between X and Y.) To show
that there is no general rule, we just need one counterexample.
Suppose that X(a, b), X(b, c), and X(c, a) are true, and that X is
false
for all other pairs. This relation is antisymmetric because we never
have
both X(u, v) and X(v, u) true at all, much less when u ≠ v. But
Y(a, b) is true (by a chain of one link) and Y(b, a) is also true
(by a chain of two links from b to c to a). Since a ≠ b, Y is
not
antisymmetric.
Again, while most of you recognized that you had to prove Z to be
reflexive, symmetric, and transitive, few of you had valid proofs of
these facts. Many of you assumed that if Y had these properties, Z
must as well, which is true but requires proof. Many of you didn't
even recognize any difference between Y and Z, just asserting that Y
had the desired properties.
Z is reflexive because if u is an arbitrary dog, Z(u, u) is defined
to be "(u = u) ∨ Y(u, u)", and this is true whether Y(u, u) is
true or not. (A fair number of you had valid proofs for this.)
To prove that Z is antisymmetric, we let u and v be arbitrary dogs and assume
that Z(u, v) and Z(v, u) are both true. We have the assumption
that Y is
antisymmetric, and we want to prove that u = v. Z(u, v) is defined as
"(u = v) ∨ Y(u, v)". If u = v is true, we are done because that
is our goal. If u ≠ v is true, then Y(u, v) must be true.
Similarly,
since Z(v, u) is true, if u ≠ v then Y(v, u) is true. By the
antisymmetry of Y, Y(u, v) and Y(v, u) together imply u = v. So
our
assumptions give us u = v in either case, and we have proved the
antisymmetry of Z.
To prove that Z is transitive, we let u, v, and w be arbitrary dogs
and assume both Z(u, v) and Z(v, w). We want to prove Z(u, w), and
part (b) of this problem tells us that Y is transitive. We have
several cases based on whether u = v and whether v = w. Case 1 (u
= v): in this case Z(u, w) follows from Z(v, w) by substitution. Case 2 (v = w):
now Z(u, w) follows from Z(u, v) by substitution. Case 3 (u ≠ v
∧ v ≠ w): Now Z(u, v) can only be true because of Y(u, v),
and Z(v, w) can only be true because of Y(v, w). So by
transitivity of Y, Y(u, w) is true and by definition of Z this
makes Z(u, w) true. Since we have proved our conclusion in each
case and at least one of the three cases must be true, we have
proved that Z is transitive.
The Hasse diagram for the dog set of Question 2 has four points,
one each for a, b, c, and d. There is an inverted V shape including
a, c, and d, with a above c and d (since c and d each have a path
upward to a). The point for b has no edges to any other points, and can be put anywhere.
The Chinese Remainder Theorem says that when two periodic processes have relatively prime periods x and y, the combined process has a period of xy. In this case, the two processes synchronize every 24(35) = 840 days.
I will next run out of canned food in 24 - 6 = 18 days, and I will
next run out of dry food in 35 - 13 = 22 days. The number x of days
until I next run out of both satisfies the two congruences x ≡
18 (mod 24) and x ≡ 22 (mod 35). The Chinese Remainder
Theorem says that this pair of congruences is equivalent to the
single congruence x ≡ (18)(11)(35) + (22)(-16)(24) (mod 840),
using the coefficients found in part (b). This evaluates to 6930 -
8448 = -1518. To get the actual number of days we need to find the
first positive natural that is congruent to this number mod 840.
Adding 840's we get -678, then 162, which is the answer.
Several people got the correct answer by computing the days on
which I run out of canned food (18, 42, 66, 90, 114, 138, 162,
186,...) and the days on which I run out of dry food (22, 57, 92,
127, 162, 197,...) and noting that 162 is the first number on both
lists. This was an effective method if you were able to
carry out about ten two-digit or three-digit additions all
correctly. Of course you were luck the common solution came up
early in the progress toward 840 -- it could have come anywhere in
that interval.
We have argued that multiplying by z, modulo m, is a bijection if z
and m are relatively prime. (This is the reason that a hash function
that jumps z entries for every collision will eventually try all
entries.)
So if g(n) = 24n % 35, then g is a bijection. Our function f is the
composition of this g with another function h(n) = n + 9, so that f(n)
= h(g(n)). The function h is also a bijection because it has an
inverse -- the function h-1(n) = n - 9 undoes h. (The fact
that 9 is relatively prime to 35 is irrelevant -- the functions n + 7
and n - 7 are also inverses of each other.) So we are done because
the composition of two bijections is a bijection.
We can also prove f to be a bijection just by computing its
inverse, because only bijections can have inverses. If we let n be
arbitrary and let y = f(n) = (24n + 9) % 35, we can get the inverse
of f by solving for n in terms of y. We have that y - 9 ≡
24n (mod 35), and we computed in part (b) that -16 is a
multiplicative inverse of 24, modulo 35. So we have that -16(y -
9) ≡ -16(24n) ≡ n (mod 35). Letting g(y) be the
function -16(y - 9), we have that g(f(n)) = -16(24n + 9 - 9)
≡ n (mod 35) and that f(g(y)) = 24(-16(y - 9)) + 9 ≡ y
- 9 + 9 ≡ y (mod 35). So f and g are inverses, and f must be
a bijection.
Last modified 7 March 2012