AC02 Calc

ANALYTIC COMBINATORICS
PART TWO
Analytic
Combinatorics
Philippe Flajolet and

Robert Sedgewick
2. Labelled structures and
EGFs
http://ac.cs.princeton.edu
Analytic combinatorics overview
A. SYMBOLIC METHOD
1. OGFs
■f 2. EGFs
3. MGFs
B. COMPLEX ASYMPTOTICS
specification
SYMBOLIC METHOD
4. Rational & Meromorphic
5. Applications of R&M
6. Singularity Analysis
7. Applications of SA
8. Saddle point
Analytic
Combinatorics

Robert Sedgewick
GF
equation
Attention : Much of this lecture is a quick review of material in Analytic

Combinatorics, Part I
One consequence: it is a bit longer than usual
To: Students who took Analytic Combinatorics, Part I

Bored because you understand it all?
GREAT! Skip to the section on labelled trees and do the exercises.
To: Students starting with Analytic Combinatorics, Part II
Moving too fast? Want to see details and motivating applications?
No problem, watch Lectures 5, 7, and 9 in Part I.
PART TWO
Analytic
Combinatorics

Robert Sedgewick
2. Labelled structures and EGFs

• Basics
• Symbolic method for labelled classes
• Words and strings
• Labelled trees
• Mappings
II.2a.EGFs.Basics
Labelled combinatorial classes
have objects composed of N atoms, labelled with the integers 1 through N.

Ex. Different unlabelled objects
Ex. Different labelled objects
Labelled class example: cycles

Q. How many cycles of labelled atoms?
y 3 = 2
A. (/V-l)!
Labelled class example 2: pairs of cycles
Q. How many unordered pairs of labeled cycles of size N ?
X 4 = 1 1
stay tuned (next lecture)
Basic definitions (labelled classes)
Def. A set of N atoms is said to be labelled if they can be distinguished from one
another.
Wlog, we use labels 1 through N to refer to them.
Def. A labelled combinatorial class is a set of combinatorial objects built from

labelled atoms
and an associated size function.
Def. The exponential generating function (EGF) associated

with a labelled class is the formal power series A(z) =
object name
a£A
^\a\ <—size function

class name
Fundamental (elementary) identity
aeA
r N
N> 0
N\
Q. How many objects of size N ?
A. Am = N\[z n ]A(z)
With the symbolic method, we specify the class and at the same time characterize
the EGF
Basic labelled class 1: urns
Def. An urn is a set of labelled atoms.
Ua= 1
counting sequence ECF
U N = 1 e z
N>0
N\
e
Basic labelled class 2: permutations
Def. A permutation is a sequence of labelled atoms.
Pi = 1
©
©
(TV
)© ©
(T)(
)© ©
@@ @(
)© ©
©® ©(
)© ©
P 2 = 2 © (
)© ©
©
® (
)® ©
Pb
6 ©
T
T
P 4 = 24
counting sequence
Pn = N\
N>0
N\z n
Nl
E z
N>0
ECF
1 - Z
" 1 -Z
10
Basic labelled class 3: cycles
Def. A cycle is a cyclic sequence of labelled atoms
counting sequence ECF
Y n = (N- 1)! In—!—
1 -z
N> 1
(N — 1 )\z N
N!
N> 1
111
1 -Z
y 4 = 6
Labelled ("star") product operation for labelled classes
is the analog to the Cartesian product for unlabelled classes
Def. Given two labelled combinatorial classes A and B, their labelled product A*B
is a set of
ordered pairs of copies of objects, one from A and one from B, relabelled in all
consistent ways.
12
Labelled ("star") product operation for labelled classes
Ex. 2. A permutation of length N is a star product of N atoms
Z+Z+Z+Z
Combinatorial construction for permutations: P= Z N
PART TWO
Analytic
Combinatorics

Robert Sedgewick
• Basics
• Labelled trees
• Mappings
II.2a.EGFs.Basics
Analytic
Combinatorics

Robert Sedgewick
PART TWO
. Labelled structures and EGFs

• Labelled trees
• Mappings
II.2b.EGFs.Symbolic
Combinatorial constructions for labelled classes
construction
notation
semantics
disjoint union
A + B
disjoint copies of objects from A and B
labelled product
A* B
ordered pairs of copies of objects,

one from A and one from B
relabelled in all consistent ways
sequence
SEQ(A)
sequences of objects from A
set
SET( A)
sets of objects from A
cycle
CYC (A)
cyclic sequences of objects from A
A and B are
combinatorial classes
of labelled objects
16
The symbolic method for labelled classes (transfer theorem)
Theorem. Let A and B be combinatorial classes of labelled objects with EGFs A(z )
and B(z). Then
construction
notation
semantics
ECF
disjoint union
A + B
A(z) + B(z)
labelled product
A* B

A(z)B(z)
SEQk(A) or A k
k- sequences of objects from A
A(z) k
sequence
SEQ(A)
1 - A(z)
set
SETk(A)
/c-sets of objects from A
A(z) k /k\
SET (A)
e A(z)
CYCk(A)
/(-cycles of objects from A
A(z) k /k
cycle
CYC (A)
cycles of objects from A
In ’
1 -A( z)
17
In-class exercise
Check the star-product transfer theorem for a small example.
A(z) = lo|y = T2 = 5 ( z ) c ( z ) /
18
The symbolic method for labelled classes: basic constructions
urns
cycles permutations
construction U = SET(Z) Y=CYC(Z ) P=SEQ(Z)
EGF
U(z ) = e z
Y(z) = In ——
V ; 1 -z
counting
sequence
U N = 1
Y n = (N- 1)! P n = N!
construction
notation
EGF
disjoint
union
A + B
A(z) + B(z)
labelled
product
Air B
A( z )B(z)
SEQk (A )
Mzf
sequence
SEQ ( A )
1 -A(z)
set
SET* (A )
A(z) k /k\
SET(A)
e A(z)
CYCk (A )
A(z) k /k
cycle
CYC (A)
ln l -«Z)
19
Proofs of transfers
are immediate from GF counting
A + B
A(z) + B(z)
A* B
7 eAxB
aceA
f\a\ 4- \/3\\ zl“l+l^l

V \ a \ ) (M + 1^1)!
=mm
20
Proofs of transfers
are immediate from CF counting
Z N Z N Z N
A(z) k = ^{#/c-sequences of size N} — = ^ cycles of size N} — = ^ ^ ! {#^' sets of
size N} —
n >o ' n>o ' N>0
A( 7 )k 7 N A(7) k 7 N
V cles of size N f JA. kl = X^ #/c ~ sets of size /syi
N> 0 ' ' iV>0
class
construction
EGF
k-sequence
SEQ k (A )
A(z) k
sequence
k-cycle
cycle
SEQ k (A ) = SEQo(A ) + SEQi (A ) + SEQ 2 (A ) + . . .
CYCki A )
CYC k (A ) = cyc 0 (a ) + crCiM) + crc 2 M) + ...
1 + /\(z)+/\(z) 2 + /\(z) 3 + .
, . ^( z ) . ^( z ) 2 . A (z) 3 ,
"" 1 -*00
-In 1
1+ 1 + 2 + 3 +
1 -A(z)
k-set
SET k (A )
k\
set
SETk(A ) = SETo(A ) + SETi (A ) + SET 2 (/\ ) + . . .
, . >U Z ) , ^( z ) 2 , ^( z ) 3
1! 2! 3!
CD
II
21
A standard paradigm for analytic combinatorics
Fundamental constructs
•elementary or trivial
•confirm intuition
Compound constructs
• many possibilities
•classical combinatorial objects
•expose underlying structure
Variations
•unlimited possibilities
• not easily analyzed otherwise
22
A combinatorial bijection
[from AC Part I Lecture 5]
A permutation is a set of cycles.
Standard representation
Set of cycles representation
23
Enumerating permutations
How many permutations of length N ?
Construction
P = SEQ(Z) “A permutation is a sequence of labelled atoms"
EGF equation
II
Counting sequence
P N = N\[z n ]P(z) = N\
How many sets of cycles of length N ?
Construction
p* _ SET (CYC (Z)) A P ermutation is a set of cycles"
EGF equation
P (z) exp (in 1 _ z ) - 1 _ z
Counting sequence
= N![z n ]P*(z) = N!
24
Derangements
A group of N graduating seniors each throw their hats in the air akroom.
and each catch a random hat.
Q. What is the probability that nobody gets their own hat back ?
Definition. A derangement is a permutation with no singleton cycles
25
Enumerating derangements
How many permutations of length A/?
Construction
EGF equation
Counting sequence
p*
P*(z)
p*
' N
SET (CYC (Z))

exp (In--)
z 7 1 -z
N\[z N ]P* (z) = N\
n*
“A permutation is a set of cycles"
How many derangements of length N ?
Construction
EGF equation
o = 5£r(cyc >1 (z))
D(z)
lZ 2 /2+z 3 /3+z 4 /4+... _
“Derangements are permutations

with no singleton cycles"
Expansion [ zN ]D( z ) =
D
N\
0<k<N
= exp (in - - z) = —
(—1 ) k 1
k\
26
Derangements
A group of N graduating seniors each throw their hats in the air

and each catch a random hat.
Q. What is the probability that nobody gets their own hat back ?
A. -= 0.36788
e
27
More variations on the theme
[from AC Part I Lectures 5 and 7]

How many permutations of length N have no cycles of length < M (generalized
derangements )?
Construction
OGF equation
D m = SET(CYC >m (Z))
“Derangements are permutations

whose cycle lengths are all > M "
Z M +1 , Z M+2
D m (z) = e M +i + M + 2
1 -z
How many permutations of length N have no cycles of length > 2 ( involutions )?
Construction
l = SET(CYQ a (Z))
“Involutions are permutations

whose cycle lengths are all 1 or 2"
OGF equation
l(z)
J+Z/2
28
Standard paradigm example: permutations
DERANGEMENTS
(no singleton cycle)
D = SET (CYC>i( Z ))
GENERALIZED DERANGEMENTS
(all cycle lengths > r)
D> r = SET (CYC< r ( Z))
e -z-J /2-...-Z/r
PERMUTATIONS
with M cycles
P m = SET m (CYC( Z ))
GENERALIZED INVOLUTIONS
(no cycle length > r)
l< r = SET (CYC< r (Z))
l< r (z ) = e z+z2 /2+-+z f A
PERMUTATIONS
with arbitrary
cycle length constraints
P n = SETq(CYC( Z ))
Pq ( z ) = ! k
Ml v 1
PERMUTATIONS
P= SET (CYC( Z ))
P(z) = e
In
= p 1 - z —
1
1 -z
INVOLUTIONS
(cycle lengths 1 or 2)
1= SET(CYCi, 2 (Z))
/(z) = e z+z2 / 2
29
Analytic
Combinatorics

Robert Sedgewick
PART TWO
. Labelled structures and EGFs
• Labelled trees
• Mappings
II.2b.EGFs.Symbolic
PART TWO
Analytic
Combinatorics

Robert Sedgewick
CAMBRIDGE
• Labelled trees
• Mappings
II.2c.EGFs.Words
Words and strings
A string is a sequence of N characters (from an M-char alphabet). There are M N

strings.
Correspondence
• For each / in the /<th set in the word set the /th char in the string to k.
• If the /th char in the string is k, put / into the /cth set in the word.
Q. What is the difference between strings and words?
A word is a sequence of M labelled sets (having N objects in total). There are

M^words.
123456789
Typical string
2 4 2 4 5 5 1 2
Typical word
{7}{183}{}{24}{569
A. Only the point of view (sequence of characters vs. sets of indices).
32
Balls and urns
Throw N balls into M urns, one at a time.
®©@®©©©©©
N= 9
1 2 3 4 5 «- M = 5
Balls-and-urns sequences are equivalent to strings and words
Corresponding string 251511443
Corresponding word {356}{1}{9}{78}{24}
33
Words
Def. A word is a sequence of M urns holding N objects in total.

Q. How many words ?
“throw N balls into M urns”
Class Wm, the class of M -sequences of urns
Size | w\, the number of objects in w
Z M Z N
ECF W M (Z) = E ^7 = E Wmn W
w^Wm N>0
Construction Wm = SEQm(SET(Z))
OGF equation Wm(z) = (e z ) M = e Mz
Counting sequence N\[z n ]Wm(z) = M N
Atom
type
class
size
CF
labelled atom
Example {7}{183}{}{24}{569}
^ ® . C® ©
u © © 0 © © ©
34
Strings and Words (summary)
class type CF type
example
AC enumeration
prototypical
AofA application
S — SEQ(Zi + ... + Z/w)
STRING unlabelled
OGF
2 4 2 4 5 5 1 2 5
5(z) = - string search
v ; 1 - Mz
Smn = M N
( 7 ) ® (b ® ®
® ®@ * @ ®®
{7}{183}{}{24}{569}
Wm = SEQm(SET(Z)
WORD labelled EGF 242455125 W M (z) = e Mz hashing
Wmn = M N
35
Variations on words: occupancy restrictions
Def. A birthday sequence is a word where no letter appears twice.
9 7 1 3 4 8 6
Def. A coupon collector sequence is a word where every letter appears at least
once.
Birthday sequences (M-words with no duplicates)
Def. A birthday sequence is a word where no set has more than one element.
a string with no duplicate letters
Q. How many birthday sequences?
Example
{ 3 } { } { 5 } { 1 } { } { } { 4 } { 2 } { }
4 8 17 3
Class Bm, the class of birthday sequences
ECF B m (z) = ^2
w££>m
z\ w \
\w\\
E*
N> 0
r N
MN
N\
Construction B^ — SEQm(E + Z)
EGF equation Bm(z) = (1 + z)
Counting sequence
N\[z n }B m (z) = N\
Ml
(M-N)l
M(M - 1)... (M - N + 1)
37
Coupon collector sequences (M-words with no empty sets)
Def. A coupon collector sequence is an M-word with no empty set.
Q. How many coupon collector sequences?
a string that uses all the letters in the alphabet
Class Rm, the class of coupon collector sequences
Z M Z N
ECF Rm{z) = 2 ^ jZfT = 2 ? Rmn W
w<eRm N> 0
Example ( M = 26 )
the quick brown fox jumps over the lazy dog
Example ( M = 5 )
2 4 2 4 5 5 1 5 3
{7}{13}{9}{24}{568}
Construction R M = SEQ M (SET >0 (Z))
EGF equation ^/w( z ) = (c z — *\) M
38
Surjections
Def. An M-surjection is an M-word with no empty set.
Alt name for "coupon collector sequence"
Def. A surjection is a word that is an M-surjection for some M.
Q. How many surjections of length N ?
Class Rm, the class of M-surjections

Construction
R m = SEQ m (SET > 0 (Z))
EGF equation
Rm(z) = 0 Z - 1 )
Coefficients
MN
Class R, the class of surjections
Construction
R = SEQ(SET > 0 (Z))
EGF equation
R(z)
1 - (e z - 1)
Coefficients
r N
N\[z N ]R(z)
N\
2(ln2) N + 1
R] = 1
1 1
1 2
2 1
R 2 = 3
1
2
13
Best handled with

complex asymptotics
(stay tuned)
39
Some variations on words
M- SURJECTIONS
(M-word, all letters used)
Rm = SEQm(SET>o(Z)
R M (z) = (e z - 1) M
Generalized Birthday
MAX occupancy M-WORDS
(all letter counts < b)
Wm = SEQ M (SET< b (Z)
= (E k < b z k /k\) M
SURJECTIONS
(M-word for some M, all letters used)

R = SEQ(SET>o(Z)
M-WORD
Wm = SEQm (SET(Z)
W M {z) = (e z ) M = e Mz
Generalized Coupon Collector
MIN occupancy M-WORDS

(all letter counts > b)
Wm = SEQm (SET>ti Z)
wtf(z) = X k>b z k /k\)
OCCUPANCY CONSTRAINED M-WORDS

(arbitrary letter count constraints)
W M ci = SEQm (SETn(Z)
W M a(z) = (E teQ z*/«) M
40
PART TWO
Analytic
Combinatorics

Robert Sedgewick
CAMBRIDGE
• Labelled trees
• Mappings
II.2c.EGFs.Words
PART TWO
Analytic
Combinatorics
Robert Sedgewick
CAMBRIDGE
• Labelled trees
• Mappings
II.2d.EGFs.Trees
Labelled trees
Def. A labelled tree with N nodes is a tree whose nodes are labelled with the
integers 1 to N.
Q. How many different labelled trees of size N ?

Q. Order of subtrees significant?
Q. Rooted?
Q. Binary?
Q. Labels increase along paths?
Some of these questions are trivial; others are classic.
All of them are easily answered with analytic combinatorics.
43
Counting labelled trees
same trees
class
rooted ordered
rooted unordered (Cayley)
unrooted unordered
increasing Cayley
increasing binary
reason
different trees
reason
order of
subtrees is
significant
order of
subtrees is not
significant
same labels on
middle node
root label
different labels
on middle node
different labels
on paths
order of
subtrees is
significant
44
Labelled trees
A. N\ Cn . Proof. Label any canonical walk of every unlabelled tree N\ different

ways
45
Labeled rooted ordered trees
Q. How many different labelled rooted ordered trees of size N ?
Class L, the class of labelled rooted ordered trees
7 |/| yN
ecf 4z) = Ew=ES
leL 11 ‘ N> o
Construction
EGF equation
Counting sequence
L = Z'k SEQ{Z) “A tree is a root and a sequence of trees"
Liz) = -—— < - Same as OGF for unlabelled trees

K ’ 1 - L(z)
Ln = N\[z N ]L(z) = N![z N ]C(z) = N\Gm + - a/! ways to label a tree walk
_ 1 (2N-2\ _ (2N-2)!
'N\N- 1 ) (N — 1)!
Stirling's approximation
Cayley trees
Q. How many different labelled rooted unordered trees of size N ?
< -24 ways to label
◄-1 2 ways to label
< -24 ways to label
T 4 = 64
4 ways to label
A. N N ~\ Proof. Stay tuned: Cayley trees are special cases of mappings (next
section)
47
Increasing Cayley trees

Q. How many different Cayley trees of size N with increasing labels on every path ?
"Cayley" = "rooted, labelled, unordered"
48
Increasing binary trees
Q. How many different binary trees of size N with increasing labels on every path ?
Boxed product construction for labelled classes
construction
boxed product
Example
notation
semantics
A = B D ★ C
subset of B^C where smallest

labelled element is from B
50
Transfer theorem for the boxed product
construction
boxed product
notation
A= B a ★ C
semantics
subset of B^C where smallest

labelled element is from B
ECF
A'(z) = B'(z)C(z)
Proof.
An —
= E
1 <k<N
(N — 1)!
- E
'N - 1
k- 1
B,
^C
k^N—k
N-k
1 <k<N
(it- 1)! (N — k)\
A\z) = ttEttZ N -'
N>1
(N-1)!
E E
N>1 1 <k<N
Bi
^N-k N -1
(it - 1)! (N — /c)!
B k C,
B, C
k> 1 N>0
(it - 1)! N!
N^n+/c-1 _
= E
/c>1 N>/c
B k
(it - 1)! (A/ — it)!
N-/c N-l
k-
’E
Qv AV
(it - 1)! ^ N!
fc>1 v ’ N> 0
^'(z)C(z)
51
In-class exercise
Check the boxed-product transfer theorem for a small example.
52
Increasing trees
Class Q, the class of Cayley trees whose labels increase on every path
Construction
Q = Z D ★ SET (Q)
EGF equation
Q'(z) = e Q < 2 >
Solution
Q( z ) - in 1 _ z
Counting sequence
Qn = N![z n ]Q(z) = (N - 1)! "Cayley"
Class B, the class of binary trees whose labels increase on every path
Construction
B=E+Z D +B+B
EGF equation
B'(z) = B(z) 2
Solution
e «=1-z
Counting sequence
B n = N\[z n ]B(z) = N\
"Cayley" = "rooted, labelled, unordered"
binary" = "ordered, each node with 0 or 2 children
53
A permutation is an increasing binary tree
Increasing binary tree

representation
Standard
representation
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
54
Some variations on labelled trees
BINARY
M-ARY
ROOTED ORDERED
L = Z ★ SEQ(L)
INCREASING CAYLEY
Q=Z n ★ SET(Q)
Q(z) = In
ROOTED UNORDERED
(Cayley)
C=Z ★ SET(C)
C(z) = ze C( ^
INCREASING BINARY
8= E+Z D + B+B
B(z)
^ -z
UNROOTED UNORDERED
55
PART TWO
Analytic
Combinatorics

Robert Sedgewick
CAMBRIDGE
• Labelled trees
• Mappings
II.2d.EGFs.Trees
PART TWO
Analytic
Combinatorics

Robert Sedgewick
CAMBRIDGE
• Labelled trees
• Mappings
II.2e.EGFs.Mappings
Mappings
Q. How many A/-words of length N ?
Mi = 1
1 1
1 2
2 1
2 2
M 2 = 4
111 211 B 1 1
112 212 B 1 2
113 213 313

121 221 321
122 222 322
123 223 323
131 231 331
132 232 332
1 3 3 2 3 3 3 3 3
Mb = 27
A. N N
1111 2111
1112 2112
1113 2113
1114 2114
1121 2121
1122 2122
1123 2123
1124 2124
1131 2131
1132 2132
1133 2133
1134 2134
1141 2141
1142 2142
1143 2143
1144 2144
1211 2211
M 4 =
111 4111
112 4 112
113 4 113
114 4 114
12 1 4 12 1
1 2 2 4 1 2 2
1 2 3 4 1 2 3
1 2 4 4 1 2 4
13 1 4 13 1
1 3 2 4 1 3 2
1 3 3 4 1 3 3
1 3 4 4 1 3 4
14 1 4 14 1
1 4 2 4 1 4 2
1 4 3 4 1 4 3
1 4 4 4 1 4 4
2 11 4 2 11
3
3
64
Mappings
Def. A mapping is a function from the set of integers from 1 to N onto itself.
Example
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
32 33 34 35 36 37
9 12 29 33 5 20 30 37 26 20 13 8 2 33 29 2 35 37 33 9 35 21 18 2 25 1 20 33 23 18
29 5 5 9 11 5 11
Every mapping corresponds to a digraph
• N vertices, N edges
• Outdegrees: all 1
• Indegrees: between 0 and N
Natural questions about random mappings
• Probability that the digraph is connected ?
• How many connected components ?
• How many nodes are on cycles ?
59
Mappings
Q. How many mappings of length N ?

M 2 = 4
(inn
12 3 to to to
113 ^
121 Q
122 V ft
133 A ^
2 2 3 O
3 2 3
n; ©0
2 11
212 /rCX
2 3 2 tob
311 y
3 3 1 A
3 3 2 ^
112 Q
1 3 1 Q
221 Y
2 3 3 O
3 13 f
322 n
ill K
Mb = 27
A. N N , by correspondence with A/-words, but internal structure is of interest.

60
oaO
Lagrange inversion
is a classic method for computing a functional inverse.
Def. The inverse of a function f(u) = z is the function u = g(z).
Ex. f(u) = -- g(z) = ———
w 1 -u 1 +z
Lagrange Inversion Theorem.
If a GF g(z) = ^g n z n satisfies the equation z= f(g(z ))
n>i
with f (0) = 0 and f' (0)^0 then g n = - [a
n — 1
J(u)J ■
Proof. Omitted (best understood via complex analysis).
Ex. f( u ) =—— g n = -[ u n '](1 — u) n = (—1 ) n_1

1—u n v 7
V(-1) n z" =
n> 1
1 +Z
Analytic combinatorics context: A widely applicable analytic transfer theorem
Lagrange-Burmann inversion
A more general (and more useful) formuation:
Lagrange Inversion Theorem (Burmann form).
If a GF g(z) = ^g n z n satisfies the equation z= f(g(z))
n> 1
with f( 0) = 0 and f ' (0) =£ 0 then, for any function H(u), *
[z"]H(g(z)) = l[ U "-']H'( U )(Z) n
/-/(w) = w gives the basic theorem
One important application: enumerating mappings
62
Lagrange inversion: classic application
How many binary trees with N external nodes?
Class
Size
Construction T = Z + T x T
OGF equation T(z) = z + T(zf

If a CF g(z) = satisfies the equation z = f(g (z))
n —' 1 —i ( u \ n
with f(0) = 0 and f' (0) * 0 then gn = ~[u n ■
Take M = N and k = N - 1 in
1 = + M ~ 1 ^ 7 k
(1 -z) M \ )
z = T(z) - T(z)‘
Extract coefficients
by Lagrange inversion
with f(u) = u - u 2
^)=
1 / 2N — 2\
N\N- 1 ) /
T, the class of all binary trees

The number of external nodes
63
Cayley trees
Class
ECF
C, the class of labelled rooted unordered trees
c ( z ) = 5Z
cec
Cn
N> 0
Example
6 2 11
Construction
C = Z*(SET(C))
EGF equation
C(z) = ze c( - z ' 1
with f(u) = u/e u
[z N }C(z) = l[u N -
”1 r N—
= N [U
C N = N![z n ]C
"a tree is a root connected to a set of trees"
/ U \ N
' u/e u '
If a CF g(z) = ^g„z n satisfies the equation z= f(g (z))
n —i 1 _i / tr \ n
with f(0) = 0 and f 1 (0) * 0 then gn = ~[u n ■
/-I
C N = N\[z n ]C(z) =(n^)
N\
/
64
Connected components in mappings
Q. How many different cycles of Cayley trees of size N ?
N N y/n
A. r ^ j — ■ (see next slide)
V2N
112
13 1
2 2 1
2 3 3
3 13
3 2 2
2 11
2 12
2 3 2
3 11
3 3 1
3 3 2
1 1
2 2
3 3
2 3 1
3 12
y 3 = 17
65
Connected components in mappings
Class
EGF
Y, the class of cycles of Cayley trees
Construction
EGF equation
with f(u) = u/e u
and H(w) =ln (1/(1 -w))
Y(z)
= In
CYC(C)
1 - C(z)
[z N ]Y(z) = ^[u N '}Y^e uN
"a component is a cycle of trees"
If a CF g(z) = J2g n z n satisfies the equation z= f(g(z))
n> 1
with f{ 0) = 0 and f ' (0) * 0 then, for any function H(u),
[z"]H(g(z)) = l[u n - 1 ]H'(u)(^) n
= E
0<k<N
N k ~ 1
= E
l</c<N
(N — k)\
Y n = N\[z n ]Y(z) = N n ~'
1 <k<N
N\
N k (N-k)l
N n ~'Q(N) ~
vys
v'tn
66
Mappings
Class
M, the class of mappings
Example
EGF
= E f
m£M
r\m
m |
1 Z N
T -E M ^ N!
N>0
1 2 13 1
Construction
M = SET(CYC(C))
EGF equation
by Lagrange-Burmann
with f(u) = u/e u
and H(u) = 1/(1 -w)
[z n ]M(z) =
M(z) = exp (in -
C(z)
) = T
N [ J (1 -u) 2
N‘
V uN
= E ("-*)
k-^
0<k<N
k\
-C(z)
N k
T\
0<k<N
If a CF g(z) = £g n z" satisfies the equation z = f(g(z))
n> 1
with f (0) = 0 and f'(0) * 0 then, for any function H(u),
[z"]H(g(z)) = l[u"- 1 ]H'(u)(^) n
- E
1 <k<N
N k ~ ]
W^y.
N N
XT
67
PART TWO
Analytic
Combinatorics

Robert Sedgewick
CAMBRIDGE
• Labelled trees
• Mappings
II.2e.EGFs.Mappings
PART TWO
Analytic
Combinatorics

Robert Sedgewick
• Labelled trees
• Mappings
• Summary
II. 2f.EGFs.Summary
The symbolic method for labelled classes (transfer theorem)
Theorem. Let A and B be combinatorial classes of labelled objects with EGFs A(z )
and B(z). Then
construction
notation
semantics
ECF
disjoint union
A + B
A(z) + B(z)
labelled product
A* B
A(z)B( z)
SEQk(A) or A k
k- sequences of objects from A
A(z) k
sequence
SEQ(A)
1 - A(z)
set
SETk(A)
/c-sets of objects from A
A(z) k /k\
SET (A )
e A(z)
CYCk(A)
/(-cycles of objects from A
A(z) k /k
cycle
CYC (A)
cycles of objects from A
In ’
1 — A(z)
boxed product
A= B a ★ C
subset of B^C where smallest labelled

element is from B
A'( z) = B'(z)C(
Constructions for labelled objects (summary)
class
construction
EGF
urns
U= SET(Z)
U(z) = e z
cycles
Y= CYC(Z)
Y( z ) - In l _ z
permutations
P = SEQ(Z)
derangements
D = SET (CYC>](Z))
D(z) = z
involutions
1 = SET (CYOAZ))
/(z) = e z+z2 / 2
words
Wm = SEQm(SET(Z))
W M (z) = e Mz
surjections
R = SEQ (SET>o(Z))
R(z) = 1
trees
L = Z+ SEQ(L)
L(z) - In 1 _ z
Cayley trees
C=Z+ SET(C)
C(z) = ze c ( z )
increasing Cayley trees
Q = Z° ★ SET ( Q )
Q'(z ) = e Q ^
mappings
M = SET (CYC(C))
M '< z > = , - C(z)
71
Analytic combinatorics overview
To analyze properties of a large combinatorial structure:
1. Use the symbolic method
• Define a class of combinatorial objects.
• Define a notion of size (and associated generating function)
• Use standard operations to develop a specification of the structure.

Result: A direct derivation of a GF equation (implicit or explicit).
Analytic
Combinatorics
Philippe Flajoletand
Robert Sedgewick
Important note: GF equations vary widely in nature
U{z ) = e z Q'(z) = e Q(z)

1
D(z)
1 -z
*(z) =
P(z) =
V(z) = In
1 -z
1 -z
M(z)
W^ b (z) = ( 1 + z + zV2! + ...+z>!)
2 - e 2
2 /?!
/(z) = e
— p z +z 2 /2
l<r(z ) = e
L{z) = In
— p z +z 2 /2+...+z r /r
1
1
1 - C(z)
Wiw(z) = e
Mz
D>r(z) =
-z-z 2 /r
1 -Z
1 -Z
C(z) = ze c ^
2. Use complex asymptotics to estimate growth of coefficients (stay tuned).
72
Direct advantages of the symbolic method
We can automate the transfer from specifications to GFs.
Ref: Automatic average-case analysis of algorithms.
by Philippe Flajolet, Bruno Salvy, and Paul Zimmerman (TCS 1991).
specification
We can use specifications to generate random structures.
Approach 1: Use a recursive program based on the specification.

Drawback: Requires quadratic time (not useful for large structures).
Approach 2: Use a probabilistic recursive program based on the specification.

Need to settle for approximate size N.
Can generate large structures in linear time.
Ref: Boltzmann samplers for random generation of combinatorial structures.
by Philippe Duchon, Philippe Flajolet, Guy Louchard and Gilles Schaefer (CPC 2004).
73
French mathematicians on the utility of GFs (continued)
"This approach eliminates virtually all calculations."
— Dominique Foata & Marco Schutzenberger, 1970
74
PART TWO
Analytic
Combinatorics

Robert Sedgewick
• Labelled trees
• Mappings
• Summary
II. 2f.EGFs.Summary
PART TWO
Analytic
Combinatorics

Robert Sedgewick
• Labelled trees
• Mappings
• Exercises
II.2g.EGFs.Exercises
Note 11.11
Ehrenfest model
Analytic
Combinatorics

Robert Sedgewick
> 11.11. Balls switching chambers: the Ehrenfesi model. Consider a system of two
cham¬
bers A and B (also classically called “urns”). There are N distinguishable balls,
and, initially,
chamber A contains them all. At any instant j, j ,..one ball is allowed to change
from one
chamber to the other. Let E„ 1 be the number of possible evolutions that lead to
chamber A
containing £ balls at instant n and E^{z) the corresponding EGF. Then
E^\z) = ^^(coshz)^(sinhz)^ -£ , E^\z) = (cosh z) N = 2~ N (e z + e~ z ) N .
[Hint: the EGF enumerates mappings where each preimage has an even cardinality.] In
particular the probability that urn A is again full at time 2 n is
2 n N 2h
(N -2k) 2 ' 7 .
77
Note 11.31
Combinatorics of trigonometries
Analytic
Combinatorics

Robert Sedgewick
t> 11 . 31 . Combinatorics of trigonometries. Interpret tan

combinatorial classes.
tan tan z, tan(e z — 1) as EGFs of
<1
78
Assignments
1. Read pages 95-1 49 (Labelled Structures and ECFs) in text.
2. Write up solutions to Notes II.1 1 and 11.31.
Analytic
Combinatorics

Robert Sedgewick
3. Programming exercise (Extra Credit).
Program II.1. Write a program to simulate the Ehrenfest mode

(see Note II.1 1) and use it to plot the distribution of the number of
balls in urn A after 1 0 3 , 1 0 4 and 1 0 5 steps when starting with 1 0 3
balls in urn A and none in urn B.
79
PART TWO
Analytic
Combinatorics

Robert Sedgewick
• Labelled trees
• Mappings
• Exercises
II.2g.EGFs.Exercises
PART TWO
Analytic
Combinatorics

Robert Sedgewick
2. Labelled structures and
EGFs

AC02 Calc

Diunggah oleh

Informasi Dokumen

Hak Cipta

Format Tersedia

Bagikan dokumen Ini

Bagikan atau Tanam Dokumen

Opsi Berbagi

Apakah menurut Anda dokumen ini bermanfaat?

Apakah konten ini tidak pantas?

Hak Cipta:

Format Tersedia

AC02 Calc

Diunggah oleh

Hak Cipta:

Format Tersedia

ANALYTIC COMBINATORICS

Philippe Flajolet and

2. Labelled structures and

Analytic combinatorics overview

4. Rational & Meromorphic

Philippe Flajolet and

Attention : Much of this lecture is a quick review of material in Analytic

One consequence: it is a bit longer than usual

To: Students who took Analytic Combinatorics, Part I

GREAT! Skip to the section on labelled trees and do the exercises.

To: Students starting with Analytic Combinatorics, Part II

Moving too fast? Want to see details and motivating applications?

No problem, watch Lectures 5, 7, and 9 in Part I.

Philippe Flajolet and

2. Labelled structures and EGFs

• Symbolic method for labelled classes

• Words and strings

Labelled combinatorial classes

have objects composed of N atoms, labelled with the integers 1 through N.

Ex. Different labelled objects

Labelled class example: cycles

Labelled class example 2: pairs of cycles

Q. How many unordered pairs of labeled cycles of size N ?

stay tuned (next lecture)

Basic definitions (labelled classes)

Def. A labelled combinatorial class is a set of combinatorial objects built from

Def. The exponential generating function (EGF) associated

^\a\ <—size function

Fundamental (elementary) identity

Q. How many objects of size N ?

Basic labelled class 1: urns

Def. An urn is a set of labelled atoms.

counting sequence ECF

Basic labelled class 2: permutations

Def. A permutation is a sequence of labelled atoms.

Basic labelled class 3: cycles

Def. A cycle is a cyclic sequence of labelled atoms

counting sequence ECF

Y n = (N- 1)! In—!—

Labelled ("star") product operation for labelled classes

is the analog to the Cartesian product for unlabelled classes

Labelled ("star") product operation for labelled classes

Ex. 2. A permutation of length N is a star product of N atoms

Philippe Flajolet and

2. Labelled structures and EGFs

• Symbolic method for labelled classes

• Words and strings

Philippe Flajolet and

. Labelled structures and EGFs

• Words and strings

Combinatorial constructions for labelled classes

disjoint copies of objects from A and B

ordered pairs of copies of objects,

sequences of objects from A

sets of objects from A

cyclic sequences of objects from A

The symbolic method for labelled classes (transfer theorem)

disjoint copies of objects from A and B

ordered pairs of copies of objects,

k- sequences of objects from A

sequences of objects from A

/c-sets of objects from A

sets of objects from A

/(-cycles of objects from A