More Counting by Mapping

Lecture 16: Nov 7 1

This Lecture

• Division rule

• Catalan number

2
 Division Rule

If a function from A to B is k-to-1,

meaning that each element in B is mapped by exactly k elements in A

then A k B

(This generalizes the Bijection Rule.)

e.g.

This is a 2-to-1 function.

So # of ears = 2 x # of people.

3
 Example 3: Two Pairs

This is something we have encountered before when we counted poker hands.

(See L2 slide 36 for details.)

Double
Count!

A: the set of two pairs

B: the set of sequences which satisfy (1)-(6).

What we did was to show that the mapping from A to B is 1-to-2,

and thus conclude that 2|A| = |B|. Then we compute |B| and then |A|. 4
 Another Chess Problem

In how many different ways can you place

two identical rooks on a chessboard so that
they do not share a row or column?

5
 Another Chess Problem

We define a mapping between configurations

to sequences (r(1), c(1), r(2), c(2)),
where r(1) and r(2) are distinct rows,
and c(1) and c(2) are distinct columns.

A ::= all sequences (r(1),c(1),r(2),c(2)) with r(1) ≠ r(2) and c(1) ≠ c(2)

B::= all valid rook configurations

(1,1,8,8) and (8,8,1,1) maps

to the same configuration.

The mapping is a 2-to-1 mapping.

6
 Another Chess Problem

A ::= all sequences (r(1),c(1),r(2),c(2)) with r(1) ≠ r(2) and c(1) ≠ c(2)

B::= all valid rook configurations

The mapping is a 2-to-1 mapping.

Using the generalized product rule to count |A|,

there are 8 choices of r(1) and c(1),
there are 7 choices of r(2) and c(2),
and so |A| = 8x8x7x7 = 3136.

Thus, total number of configurations

|B| = |A|/2 = 3136/2 = 1568.
7
 Round Table

How many ways can we seat n different people at a round table?

Two seatings are considered equivalent if one

can be obtained from the other by rotation.

equivalent

8
 Round Table

A ::= all the permutations of the people

B::= all possible seating arrangements at the round table

Map each permutation in set A to a circular seating arrangement

in set B by following the natural order in the permutation.

9
 Round Table

A ::= all the permutations of the people

B::= all possible seating arrangements at the round table

This mapping is an n-to-1 mapping.

Thus, total number of seating arrangements

|B| = |A|/n = n!/n = (n-1)!
10
 Counting Subsets

Now we can use the division rule to compute more formally.

How many size 4 subsets of {1,2,…,13}?

Let A::= permutations of {1,2,…,13}

B::= size 4 subsets

map a1 a2 a3 a4 a5… a12 a13 to {a1,a2 ,a3, a4}

(that is, take the first k elements from the permutation)

How many permutations are mapped to the same subset??

11
 Counting Subsets

map a1 a2 a3 a4 a5… a12 a13 to {a1,a2 ,a3 , a4}

a2 a4 a3 a1 a5 … a12 a13 also maps to {a1,a2 ,a3, a4}

as does a2 a4 a3 a1 a13 a12 … a5

4! 9!

Any ordering of the first four elements (4! of them),

and also any ordering of the last nine elements (9! of them)
will give the same subset.

So this mapping is 4!9!-to-1

12
 Counting Subsets

Let A::= permutations of {1,2,…,13}

B::= size 4 subsets

13!  A  4 !9 ! B
13  13!
So number of 4 element subsets is
  ::
4 4!9!

Number of m element subsets of an n element set is

n  n!
  ::
m m !(n  m)!
13
 MISSISSIPPI

How many ways to rearrange the letters in the word “MISSISSIPPI”?

Let A be the set of all permutations of n letters.

B be the set of all different words by rearranging “MISSISSIPPI”.

How many permutations are mapped to the same word?

4! possible ways to rearrange the S giving the same word

MISSISSIPPI 4! possible ways to rearrange the I giving the same word

2! possible ways to rearrange the P giving the same word

The mapping is 4!4!2!-to-1, and so there are 11!/4!4!2! different words.

14
 Example: 20 Mile Walk

I’m planning a 20-mile walk, which should include 5 northward miles, 5

eastward miles, 5 southward miles, and 5 westward miles.

How many different walks are possible?

There is a bijection between such walks and words with 5 N’s, 5 E’s, 5 S’s,
and 5 W’s.

The number of such words is equal to the number of rearrangements:

20!
5!5!5!5!

15
 Exercises

What is the coefficient of x7y9z5 in (x+y+z)21?

There are 12 people. How many ways to divide them into 3 teams, each
team with 4 people?

16
This Lecture

• Division rule

• Catalan number

17
 Monotone Path

A monotone path from (0,0) to (n,n) is a path consist of “right"

moves (x-coordinate increase by 1) and “up" moves (y-coordinate
increase by 1), starting at (0,0) and ending at (n,n).

How many possible monotone paths from (0,0) to (n,n)?

We can map a “right” move to an “x” and a “up” move to a “y”.

There is a bijection between monotone paths and words with n x’s and n y’s.
And so the answer is just
18
 Monotone Path

We say such a path “lower-right" monotone path if all of the

points (xi,yi) on the path has xi >= yi.

lower-right monotone NOT lower-right monotone

How many possible lower right monotone paths from (0,0) to (n,n)?

19
 Monotone Path

How many possible lower right monotone paths from (0,0) to (n,n)?

We can still map a “right” move to an “x” and a “up” move to a “y”.
There is a bijection between (A) lower right monotone paths and
(B) words with n x’s and n y’s, with the additional constraint that
no initial segment of the string has more Y's than X's.
The set (B) is called the set of Dyck words.
There is a bijection, but both sets look difficult to count. 20
 Parenthesis

How many valid ways to add n pairs of parentheses?

E.g. There are 5 valid ways to add 3 pairs of parentheses.

((())) (()()) (())() ()(()) ()()()

Let rn be the number of ways to add n pairs of parentheses.

A pairing is valid if and only if there are at least as many

open parentheses than close parentheses from the left.

21
 Mountain Ranges

How many “mountain ranges” can you form with n upstrokes

and n downstrokes that all stay above the original line?

/\
/\ /\ /\/\ / \
/\/\/\, /\/ \, / \/\, / \, / \

We do not know how to solve these three problems yet,

(A) lower right monotone paths, (B) Dyck words,
(C) valid parentheses, (D) mountain ranges.

But we can show that all these three problems have the same answer,
by showing that there are bijections between these sets.
22
 Parenthesis and Monotone Paths

A pairing is valid if and only if there are at least as many

open parentheses than close parentheses from the left.

(()()()) ()()()()

xxyxyxyy xyxyxyxy

We can map a “(” to an “x” and a “)” move to a “y”.

There is a bijection between (C) valid pairings and (B) Dyck words.

In slide 19, we have seen that there is a bijection between

(A) lower right monotone paths and (B) Dyck words.
So, there is a bijection between (A) and (C).
23
 Monotone Paths and Mountain Ranges

By “rotating” the images, we see that a path not crossing the diagonal
is just the same as a mountain not crossing the horizontal line.

So there is a bijection between (A) and (D). 25

 Catalan Number

Now we know that these three sets are of equal size,

although we don’t know the size.

It turns out that the answer is exactly

This is called the nth Catalan number,

and has applications in many other places as well.

One proof of this formula is by the bijective rule.

26
 Proof Plan

The number of lower right monotone paths from (0,0) to (n,n) is

The plan is to count the complement.

In slide 18, we know that the number of monotone paths is

We will show that the number of non-lower-right monotone paths is

This would imply that the number of lower-right monotone paths is

27
 Proof

The number of non-lower-right monotone paths is

The idea is to define a bijection between

(A) The set of non-lower-right monotone paths from (0,0) to (n,n)
(B) The set of monotone paths from (0,0) to (n-1,n+1)

Clearly, the size of (B) is equal to ,

as it is equal to the set of strings with 2n characters,

where n-1 of them are R (“right”) and n+1 of them are U (“up”).

So, it remains to define a bijection between (A) and (B).

28
 Bijection

Every path in (A) must “cross” the diagonal at least once.

We look at the first “crossing”, and then “flip the path”.

(picture from wiki)

In the original path, before the flipping point, we have one more U than R.
So, in the original path, after the flipping point, we have one more R than U.

In the new path, since we flip the latter part, we have two more U than R,
and thus it is a path from (0,0) to (n-1,n+1).
29
 Bijection

Every path in (A) must “cross” the diagonal at least once.

We look at the first “crossing”, and then “flip the path”.

(picture from wiki)

The easiest way to see that this is a bijection is to define the inverse function.

Given a monotone path from (0,0) to (n-1,n+1), it must cross the diagonal.
Look at the first point and flip it, we get back the non-lower-right monotone
path from (0,0) to (n,n) that was mapped to it.
30
 Catalan Number

The “reflection” technique in the proof is useful in some other problems.

Perhaps you may see it again in homeworks.

There are other proofs and other applications of Catalan number (see wiki).

31
 Quick Summary

Counting by mapping is a very useful technique.

It is also a powerful technique to solve more complicated problems.

The basic examples usually map a set into a properly defined binary strings.

Then we see how to generalize this approach by considering k-to-1 functions.

Finally we see the mapping between more complicated sets.

Anyway, the idea of define a mapping (reduction) between two problems

is probably the most important idea in computer science.

So, make sure you are familiar with it.

32