AC07-SAapps Djvu

ANALYTIC COMBINATORICS
Analytic
Combinatorics
Philippe Flajolet and

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

Robert Sedgewick
GF
equation
Analytic
Combinatorics

Robert Sedgewick
PART TWO
7. Applications of Singularity Analysis
• Simple varieties of trees
• Labelled sets
• Mappings
• Tree-like classes
II.7a.SAapps.Sets
Transfer theorem for invertible tree classes

[from Lecture 6]
Theorem. If a simple variety of trees F = Z [ x or ★] SEQ^(F)

is A-invertible where the GF satisfies F(z) = z<j>(F(z)) and
is the positive real root of 0(A) = A0'(A) then
4>'(X ) n N- 3/2
and F(z) ~ A - ^2<j>(X)/<p"(X)^ -zcf>'(\)
Important note: Singularity analysis gives both

•Coefficient asymptotics.
•Asymptotic estimate of GF near dominant singularity.
applications
general trees
binary trees
unary-binary dd
Cayley trees
[and many, many more...]
Example 1: Rooted ordered trees

Q. How many trees with N nodes?
How many trees with N nodes?
Symbolic method
Combinatorial class
Construction
OCF equation
Quadratic equation
Classic next steps
Binomial theorem
Extract coefficients
Stirling's approximation
Simplify
C, the class of all trees

C = • x SEQ(G) *
"a tree is a node and

a sequence of trees"
G(z) = z(1 + C(z) + C(z) 2 + G(z) 3 + .
G(z) — G(z) 2 = z
, 1 + v / Wi
C(z) =- - -
1 - G(z)
~ exp(2Nln(2N) - 2N + In v / 4irN - 2(Nln(N) - N + In v'lirN))
calculations
omitted
Gs= 14
Example 1: Rooted ordered trees
G, the class of rooted

ordered trees
G = Z x SEQ(G))
> f
C(z)
z
1 - C(Z)
simple variety
of trees
> r
C N ~ -i-4 n N 3 / 2
4v^
Theorem. If a simple variety of trees F = Z [ x or ★] SECMF)

is A-invertible where the GF satisfies F(z) — z0(/-(z)) and
0'(A ) N N~ 3 / 2
sfi^rWW)
0(U)
0»
0 »
1 - u
O^F
O^F
1 _ A
1 - A “ (1 - A) 2
A = 1/2
0(A) = 2
0'(A)=4
0"(A) = 16
6
Example 2: Binary trees
How many binary trees with N nodes?
Ti = 2
/■
<1
x*x
Tb = 5
Example 2: Binary trees
B, the class of binary trees

B = *x(E + B)x(E + B)
Expecting B = • + • x SEQo,2(B)? Stay tuned
> f
B(z) = z( 1 + B(z)) 2
simple variety
of trees
> f
[z n ]B(z) ~ -^4 n N 3 / 2
A/ 71

is A-invertible where the GF satisfies F(z) = z<p(F(z)) and
V2tt0"(A)/0(A)
0'(A) N N- 3 / 2
0(U) = (1 +U) 2
0» = 2(1 +u)
0 » = 2
(1+A) 2
2A(1 + A)
A = 1
0(A) = 4
0'(A) = 4
0"(A)=2
8
Example 3: Unary-binary trees
Q. How many unary-binary trees with N nodes?
M b=2
degrees of all nodes 0, 1, or 2
Ma= 4
HA A
M 5 =9
Example 3: Unary-binary trees
Asymptotics
M, the class of all

unary-binary trees
M = Z x S£Qo,i, 2 (M)
> r
M(z) = z( 1 + M(z) + M(z) 2 )
simple variety
of trees
> f
Mn
3 N n -3/2
r-sj

is A-invertible where the GF satisfies F(z) — z0(F(z)) and
0'(A ) N N~ 3 / 2
sfi^rWW)
<t>{u) = 1 + u + u 2
0'(u) = 1+2u 1 + A + A 2 = A + 2A
0 » = 2
A = 1
0(A) = 3
0'(A) = 3
0"(A)=2
10
Example 4: 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
4 ways to label
T 4 = 64
A. N N ~ ] . (See EGF lecture.)
Example 4: Cayley trees (exact, from EGF lecture)
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)) <—
"a tree is a root connected to a set of trees"
EGF equation C(z) = ze C ^
Extract coefficients i u ,,
by Lagrange inversion [z ]C(z) = [i/ ]( , )
with f(u) = u/e“ N u / e
Lagrange Inversion Theorem.
If a CF g(z ) = ^g„z n satisfies the equation z= f{g (z))
n — i f i / u \ n
with f(0) = 0 and f ' (0) * 0 then gn = ~[u n
C N = N![z n ]C(z) =(n^) /
1 r /# N-ii p uN
N n ~ 1
“ N [U Je
N!
12
C, the class of all labelled rooted

unordered trees
C = Z ★ SET ( C )
> f
C(z) = ze c
simple variety
of trees
[z N ]C(z) — ^_ e N N 3 / 2
V2ti

is A-invertible where the CF satisfies F(z) = z0(F(z)) and
0'(A ) N N- 3 / 2
v /2ji0"(A)/0(A)
0(u) = e u
0» = e u
0"(u) = e"
Ae A
A = 1
0(A) = e
0'(A) = e
0"(A) = e
13
Aside: Stirling's formula via Cayley tree enumeration
Exact, via Lagrange inversion
Approximate, via singularity analysis
Theorem. If a simple variety of trees F = Z [ x or *] SEQ$(F)

isA-invertible where the GF satisfies F(z) = z0(F(z)) and
is the positive real root of </>(A) = A<//(A) then
( W) ~
cj>(u) = e u
<P'(u) = e" e x = Ae A
<p"(u) = e u
A = 1
<t>( A) = e
^'(A) = e
0"(A)= e
Theorem.
N\ ~ a/2tiN( —) N *
Stirling's formula
14
Analytic
Combinatorics

Robert Sedgewick
PART TWO
• Labelled sets
• Mappings
II.7a.SAapps.Sets
PART TWO
Analytic
Combinatorics

Robert Sedgewick
Simple varieties of trees

Labelled sets
Mappings
Tree-like classes
CAMBRIDGE
II.7a.SAapps.Sets
Transfer theorem for exp-log labelled set classes
[from Lecture 6]
Theorem. Asymptotics of exp-log labelled sets.
Suppose that a labelled set class F = SETco(G) is exp-log(cx, |3, p)
1 1
with G(z) ~ a log- - —h ft . Then F(z ) ~
v 7 1 -z/p v 7 V 1 -z/p ]
and
[z N ]F(z)
A A
77"
r(a ) K p
(-) N
1 —a
Corollary. The expected number of C-components

in a random F-object of size N is ~cx In N.
and is concentrated there
17
Example 5: Cycles in permutations
Q. How many permutations of N elements?
Q. How many cycles in a random permutation of N elements?
Pi= 1
avg. # cycles: 1
©©©
Ps = 6
avg. # cycles: 1 .8333
Pa-24
avg. # cycles: 2.08333
18
Example 5: Cycles in permutations
P, the class of all permutations

P = SET(CYC(Z))
> f
P(z) = exp (In y—^ )
exp-log
> f
[z N ]P(z) ~ 1
# permutations: ~ N\
avg # cycles: ~lnN
cy\ r?)
Suppose that a labelled set class F = SET<t>(G) is exp-log(a, P, p)
with G(z) ~ a log
1 1
+ ft . Then F(z) ~ e^(
1 -z/p
1 - z/p
)“
and
r(a) V
( 7 ) N
1 —a
In
a log --— + p
^ - z/p
for a = 1, p = 0, and p = 1
Corollary. The expected number of G-components

Example 6: Cycles in derangements
Q. How many derangements of N elements?

Q. How many cycles in a random derangement of N elements?
D i = 0
avg. # cycles: 0
Db= 2
avg. # cycles: 1
D 4 =9
avg. # cycles: 1.33333
20
Example 6: Cycles in derangements
D, the class of all derangements

D = SET(CYC > 0 (Z))
> f
D(z) = exp (In y/yQ))
exp-log
> f
[z n ]D(z) ~ ^35
# derangements: ~ N\/e
avg # cycles: ~ In N
Suppose that a labelled set class F = SET<t>(G) is exp-log(a, P, p)
1 1
with G(z) ~ a log - - — I- /3 . Then F(z) ~ e d (
and
1 -z/p
1 -z/p
In
1 'j
y— ;(2)= a log i _ r / - + ^
z/p
for ct = 1, (5 = -1, and p = 1

21
Example 6: Cycles in generalized derangements
D, the class of all permutations

having no cycles of length W], W 2 , ... Wt
D = SET(CYC * wi (Z))
D(z) = exp(ln
rWl
1 — z Wt
rW t
w t
[z n ]D(z) = exp(-
exp-log
Wt
W t
In
1 -z
for a = 1, f3
and p = 1
1 = a log
1 -z/p
Wi
1
w t
# derangements: ~ N!/e 1 / VVl+ "' +1 / VVt

avg # cycles: ~ In N
22
Example 7: 2-regular graphs
Q. How many labelled 2-regular graphs of N elements?
undirected graphs with

all nodes degree 2
Q. How many components in a random 2-regular graph of N elements?
1 way to label 3 ways t0 | a bel
fe = 1 r 4 = 3
1 2 ways to label
Rs = 12
60 ways to label
360 ways to label

1 0 ways to label
105 ways to label
Re = 70
avg. # components:
(1 -60 + 2 ■ 1 0)/70 = 1.143
R? = 465
avg. # components:
(1 -360 + 2 • 1 05)/465 = 1.226
23
Example 7: 2-regular graphs
R, the class of 2-regular graphs
R = SET(UCYC> 2 (Z))
z z
, x ,1, 1 - - X
K z) = exp — In-
v ; PV 2 1 -z 2 4’
exp-log
[z n ]R(z)
-3/4
V 71N
e -3/4
# 2-regular graphs: ~ N\ _
-InN
avg # components:
Analytic
Combinatorics
page 1 BB
page 449
Suppose that a labelled set class F = SET<t>(G) is exp-log(cx, 3, p)
with G(z) ~ a log
1 - zip
+ /3 . Then F[z) ~ e ;3 (
1 - zip
Y
and
[z n ]F(z)
,1
—-(-) N'~ a
r(a)V
C(z) ~ a log --— + (3
1 -z/p
for a = 1 /2, (3 = 3/4, and p = 1

in a random F-object of size N is ~a In N.
PART TWO
Analytic
Combinatorics

Robert Sedgewick
Simple varieties of trees

Labelled sets
Mappings
Tree-like classes
CAMBRIDGE
II.7a.SAapps.Sets
PART TWO
Analytic
Combinatorics

Robert Sedgewick
4^nple varieties of trees

Labelled sets
Mappings
Tree-like classes
CAMBRIDGE
II.7a.SAapps.Sets
Example 7: Mappings
[from Lecture 2]
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 IB 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
27
Mappings
[from Lecture 2]
Q. How many mappings of length N ?
M 2 = 4
(inn
12 3 to to to
113 ^
121 Q
122 y r
133 A ^
2 2 3 O
3 2 3
n; ©0
2 11
212 /rCX
2 3 2 tob
3 11 b
3 3 ! X
3 3 2 LJ
112 Q
1 3 1 Q
221 Y
2 3 3 O
3 13 f
322 n
Hi A
Mb = 27
A. N N , by correspondence with A/-words, but internal structure is of interest.
28
oaO
Mapping EGFs
[from Lecture 2]
Combinatorial class
C, the class of Cayley trees < - labelled, rooted, unordered

Construction
C = Z * (557(C)) * — "a tree is a root connected to a set of trees"
EGF equation
C(z) = ze c ^
Combinatorial class
Y, the class of mapping components
Construction
y = CVC(C) < — "a mapping component is a cycle of trees"
EGF equation
V(Z) - ln 1 - C(z)
Combinatorial class
M, the class of mappings
Construction
M = SET(CYC(C)) < — "a mapping is a set of components"
1 1
ECF equation M(z) = exp(ln -—-) = -——
v 1 — C(z) 7 1 — C(z)
29
[from earlier in this lecture]

is A-invertible where the CF satisfies F(z) = z0(F(z)) and
0'(A ) N N- 3 / 2
v /2ji0"(A)/0(A)
and F(z) ~ A - ^20(A)/0"(A)0 - z0'(A)
0(A) = e
0'(A) = e
0"(A) = e
0(u) = e 1
0'(u) = e 1
0» = e 1
e A = Ae A
30
Cycles of Cayley trees
Y, the class of cycles of trees

(mapping components)
Y = CYC(C)
Y(z) — In
1 - C(z)
lln—-—
2 1 — ez
— In \fl
standard scale
[z n ]Y(z)
2 N
# cycles of trees: N\
2 N
71
2N
1 10 21 10 291231
from previous slide
C(z) ~ 1 — — ez
Stirling
N\ ~ V2kN( k — ) N
31
Mappings
M, the class of all mappings

M = SET ( Y )
\ • f \
• i | , /
>
A i i L
y(z)
r N
n n /
for ce = 1 /2, (3 = - In a/ 2, and p = 1 /e
32
Mappings overview
Theorem. If a simple variety of trees F = Z t x or *] SEQ»(F)

is A-invertible where the CF satisfies F(z) = z<t>(F{z)) and
(VlfM-
4‘W N-
I and f{z) ~ A - Cycles of Cayley trees
0 ( 0 ) = e"
0'(o) = e u
0"(o) = e u
Cayley trees: simple variety
f Asymptotics j
Y, the class of cycles of trees

(mapping components)
Y(z) = In
1 10 21 10 291231
from previous slide
Mappings
# cycles of trees: ~ f.— ~ J —N
Components: standard scale
Mappings: exp-log
33
Mapping parameters
Q. How many components in a random mapping of length N ?
Q. How many nodes on cycles in a random mapping of length N ?

i
Mi = 1
1 2
1 1
2 2
2 1
00
/W 2 = 4
avg. # components: 1.25

avg. # nodes on cycles: 1.5
123 000
113 ^
121 O'
122 Y 0
1 3 3 T U
2 2 3 O
3 2 3
Hi ©
2 12 *0
2 3 2 v>
311 v
3 3 2 6
112 Q
131 Q
2 2 1 y
2 3 3 O
3 13 f
3 2 2 6
333
Mb = 27
avg. # components: 38/27 = 1.407

avg. # nodes on cycles: 51 /27 = 1.889
34
CXX£)
Components in mappings
M, the class of all mappings

M = SET( Y)
> f
M(z) = e Y ^
y(z) ~ 1 In 1 -In Vl
2 1 — ez
exp-log
N\[z n }M(z) ~ A/!
,/V
y/2nN
n n /
avg # components:
In N
X'
• i
2 / ...
^ •
>- A
v<
/X
r-’^i \
f ‘
VTX
r * ^
• -•'V
«■
Suppose that a labelled set class F = SET®(C) is exp-log(a, |5, p)

with C(z) ~ a log - - —I- y3 . Then F(z) ~ '
and
1 -z/p
1 - z/p
Y(z)
-T^(-) N '~°
r(a) V
-ln-
2 1
ez
ct log
+ 0
for
1 - z/p
= 1 /2,0 = — ln\/2, and p = 1/e

in a random F-object of size N is ~a In N.
35
Nodes on cycles in mappings
Combinatorial class
Parameter
Construction
BGF
Expected # nodes on cycles
M, the class of mappings

the number of nodes on cycles (tree roots)
M = SET (CYC (uC))
/W(z, u) = exp (in -- - ) = --
V 1 — uC(z )/ 1 — u
N! '.Nl 9 ..M N!
)| u=1
N N j du
[Z]
uC( z )
C(z)
N Nl (1 — C(z)) 2
N! N 1 1
~ —-[z j-
N N l J 2 1 - ez
1 N!e N
nN/2
page 462
C(z) ~ 1 — \/2Vl — ez
C(z) 1 1
(1 — C(z)) 2 2 1 — ez
Stirling
N! ~ V2n N(—) N
predicted: 1 2.5
actual: 9
36
PART TWO
Analytic
Combinatorics

Robert Sedgewick
4^nple varieties of trees

Labelled sets
Mappings
Tree-like classes
CAMBRIDGE
II.7a.SAapps.Sets
PART TWO
Analytic
Combinatorics

Robert Sedgewick
• Labelled sets
• Mappings
CAMBRIDGE
II.7a.SAapps.Sets
Schema example 4: Implicit tree-like classes
Definition. A combinatorial class whose enumeration GF satisfies F(z) = <£(z, F(z))

is said to
be an implicit tree-like class with characteristic function G.
unlabelled case: number of structures is [z N }F(z)
F = CONSTRUCTS, F)
where CONSTRUCT is an arbitrary

composition of+, x, and SEQ
labelled case: number of structures is N\[z n ]F(z)

F = CONSTRUCTS, F)-
Analytic
Combinatorics
Philippe Hjjolct and

Robert Sedgewick
page 467
F(z) = (z, F(z))
\
immediate via
symbolic transfer
where CONSTRUCT is an arbitrary

composition of+, ★, SEQ, SET, and CYC
Example: Simple varieties of trees <T(z, w) = z<f>(w )
F(z) = z<j>(F{z))
39
Smooth-implicit-function tree-like classes
smooth implicit function : A technical condition that enables us to unify the

analysis of tree-like classes.
Definition. Smooth-implicit-function tree-like classes.
A tree-like class F = CONSTRUCT(F) with enumerating GF F(z) = 0(z, F(z )) is said

to be
smooth-implicit(r, s) if its characteristic function <£(z, w) satisfies the
following conditions:
• 0(z, w) is analytic at 0 and in a domain |z|< R and \w\< S for some R, S >0.
• [z N w k \<P(z, w) > 0 and >0 for some N and some k> 2, with <I>(0, 0) =*= 0.
There exist positive reals r < R and s < S such that (r, s) = s and <I> w (r, s) =
1
Example: "phylogenetic trees" Construction

[details to follow]
L = Z + SET> 2(L)
OGF equation L(z) = z + e L ^ - 1 - L(z)
Characteristic function $(z, w) = z — 1 + e w — w
(z, w) = w
O W (Z, W) = 1
"characteristic system
Characteristic system
z + e w — 1 — w — w
e w - 1 =1
solution
r= 2ln 2 — 1
s = In 2
phylogenetic trees are smooth-implicit(2ln 2 - 1, In 2)
Transfer theorem for implicit tree-like classes
Theorem. Asymptotics of implicit tree-like classes.
Suppose that F is an implicit tree-like class with characteristic function 4>(z, w)

and
aperiodic and smooth-implicit(r, s) GF F(z) = 0(z, F(z)), so that 4>(r, s) = s and
<fr w (r, s) = 1.
Then F(z) converges at z = r where it has a square-root singularity with
2r$Ar, s)
& ww (r , s)
Example: binary trees

(alternate)
Construction
B = • + • x SEQo, 2 ( B )
B(z) =z + zB{z) 2
s = 1/2
r= 1/4
OGF equation
Characteristic function
4>(z, w) — z + w
z + w 2 = w
$ z (z, w) = 1
4> w (z, w) = 2 w
$mv(z, w)=2
a — 2
Characteristic system
2w — 1
Coefficient asyptotics [z N ]6(z) ~ —4 n N 3/2
\/K
41
Example 8. Bracketings
Def. A bracketing of N items is a tree with N leaves and no unary nodes
Analytic
Combinatorics
page 69
Applications
•Parenthesizations.
•Series-parallel networks.
•Schroder’s 2nd problem
42
Example 8: Bracketings
Q. How many bracketings with N leaves?
= 1 5 2 = 1
A\
X^ \
5b = 3
All nodes of degree 0 (leaves) or >1 (internal nodes)
size: number of leaves
x^ *X*
54 = 1 1
43
Q. How many parenthesizations of N items?
Si= 1
(a b)
5z= 1
(a b c)
((a b) c) (a (b c))
5b = 3
((a
(((a b)
(abed)
b) c d) (a (b c) d) ((a b) c d)
((a (b c)) d)
c) d) (((a b) c) d)
(a C(b c) d))
((a b) (c d))
((a b c) d) (a (b c d))
5 4 = 1 1
Three additional equivalent structures.
and-or conjunctive
propositions
aA((bvc)AdA(evf)vg)A(hv(i a j)vk )a (1 vm)
series-parallel networks
<D—0>
—®—
45
S, the class of all brat

S = Z + SEQ >i(
5(z) = z +
1 - 5(z)
- 1 - 5(z)
Suppose that F is an implicit tree-like class with characteristic function <I>(z,

w) and
aperiodic and smooth-implicittr, s) CF F(z) = 4>(z, F(z)), so that <f>(r, s) = sand
<J>i»(r, s) = 1.
Then F(z) converges at z= r where it has a square-root singularity with
“ fiiv5\v
F(z) ~ s - aV'l -z/r and([ z N ]F(z) ~ (y)”AT
(where q =
2r$ z (r,s)
, s)
[z N ]S(z)
[ details left for exercise ]
Y 8s/2n K r
with r — 3 — 2 V 2
46
Example 9. Labelled hierarchies (phylogenetic trees)
Def. A labelled hierarchy of N items is a tree with N labelled leaves and no unary
nodes
Applications
•Classification.
•Evolution of genetically related organisms.

•Schroder’s 4th problem
Analytic
Combinatorics
page 1 28
47
Q. How many different labelled hierarchies of N nodes?
U = 26
48
L, the class of labelled hierarchies
L = Z + SET> 2 (L)
> f
L(z ) — z + e i( ' z ' ) — 1 — L(z)
implicit
tree-like
Suppose that F is an implicit tree-like class with characteristic function <t>(z,

w) and
aperiodic and smooth-implicit(r, s) GF F(z) = 4>(z, F(z)), so that Mr, s) = sand <J
> w (r, s) = 1.
Then F(z) converges at z= r where it has a square-root singularity with
f(z) ~ s „ V.
z + e w — 1 — w — w
e w - 1 = 1
f = 2 In 2 — 1
s = In 2
with r = 2 In 2 — 1
<£>(z, w) = z — 1 + e w — w
&z(z, W) = 1
$ w (z, w) = e w - 1
$ wvv (z, w) = e w
$ z (r,s) = 1
$mv(r,s) = 2
a — \/2 In 2 — 1
49
PART TWO
Analytic
Combinatorics

Robert Sedgewick
• Labelled sets
• Mappings
CAMBRIDGE
II.7a.SAapps.Sets
Analytic
Combinatorics

Robert Sedgewick
PART TWO
^Set scherga^>
• Mappings
• Summary
II.7e.SAapps.Summary
Singularity analysis: examples of applications
construction generating function coefficient asymptotics
rooted ordered trees
G = Z x SEQ(G )
C(Z) =1-C(Z)
'a n N 3 ' 2
40E
binary trees
B = • x (E + B) x (E + B)
B = • + • x SEQo, 2 ( B )
e(z) = z(i +e(z) 2 )
B(z) — z + zB(z) 2 )
1 4 N N } / 2
v n
unary-binary trees
M = • x SEQo,i, 2 ( M )
M(z) = z( 1 + M(z) + M(z) 2 )
' _3 n N- } ' 2
V 471 / 3
Cayley trees
C = Z ★ SET(C )
C(z) = ze C( ' z ' )
N\ L e N N~V 2 = N n -’
y/2/K
mapping components
K = CYC(C )
K<Z) - hl 1 - C(z)
~N!— - J—N n
2N V 2N
mappings
M = SET(K )
M(z) = e K(z) = , _' c( z)
e N
~ /V! _~ /V
V2^N
2-regular graphs
R = SET( UCYC>2 (Z))
„-z/2-z 2 /4
e -3/4
~ /V! _
VnN
labelled hierarchies
L = Z + SET> 2 ( L )
L(z) = z + e L ^ — 1 — L(z)
yjl 111 2 - 1 N!
2\/tiN 3 (2 In2 — 1 ) N
52
"If you can specify it, you can analyze it"
Singularity analysis is an effective approach for analytic transfer from GF

equations
to coefficient asymptotics for classes with CFs that are not meromorphic.
Schema can unify the analysis for entire families of classes.
schema
technical condition
construction
coefficient asymptotics
Labelled set
exp-log
F = SET(G)
^0 i N
( )
r(cOV
Simple variety
of trees
invertible
F = Z x SEQ( F)
F = Z ★ SEQ( F)
’(Vn - 3 / 2
y/an p
Context-free
irreducible
Family of (+, x)
constructs
L_ (Vn - 3 / 2
y/ait p
Implicit tree-like
smooth implicit
function
F = CONSTRUCT (F)
a ^C) N N~V 2
2 r '
Next: GFs with no singularities.
53
Analytic
Combinatorics

Robert Sedgewick
PART TWO
^Set scherga^>
• Mappings
• Summary
II.7e.SAapps.Summary
Analytic
Combinatorics

Robert Sedgewick
PART TWO
^Set scherga^>
• Mappings
• Exercises
II.7f.SAapps.Exercises
Web Exercise VII.1
Bracketings (Schroder's 2nd problem)
Analytic
Combinatorics

Robert Sedgewick
Web Exercise VII. 1 . Use the tree-like schema to develop an

asymptotic expression for the number of bracketings with N leaves
(see Example 1.1 5 on page 69 and Note VII.19 on page 474).
56
Assignments
1. Read pages 439-540 (Applications of Singularity Analysis) in text.
Usual caveat: Try to get a feeling for what's there, not understand every detail.
2. Write up a solutions to Web Exercise VII.1.
3. Programming exercise.
Program VII.1 . Do r- and 0-plots of the GF for bracketings

(see Web Exercise VII.1).
Analytic
Combinatorics

Robert Sedgewick
57
Analytic
Combinatorics

Robert Sedgewick
PART TWO
^Set scherga^>
• Mappings
• Exercises
II.7f.SAapps.Exercises
Analytic
Combinatorics

Robert Sedgewick
PART TWO
7. Applications of
Singularity Analysis

AC07-SAapps Djvu

Diunggah oleh

Informasi Dokumen

Judul Asli

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

AC07-SAapps Djvu

Diunggah oleh

Hak Cipta:

Format Tersedia

ANALYTIC COMBINATORICS

Philippe Flajolet and

Analytic combinatorics overview

4. Rational & Meromorphic

Philippe Flajolet and

Philippe Flajolet and

7. Applications of Singularity Analysis

• Simple varieties of trees

Transfer theorem for invertible tree classes

Theorem. If a simple variety of trees F = Z [ x or ★] SEQ^(F)

and F(z) ~ A - ^2<j>(X)/<p"(X)^ -zcf>'(\)

Important note: Singularity analysis gives both

•Asymptotic estimate of GF near dominant singularity.

[and many, many more...]

Example 1: Rooted ordered trees

How many trees with N nodes?

C, the class of all trees

"a tree is a node and

G(z) = z(1 + C(z) + C(z) 2 + G(z) 3 + .

Example 1: Rooted ordered trees

G, the class of rooted

Theorem. If a simple variety of trees F = Z [ x or ★] SECMF)

Example 2: Binary trees

How many binary trees with N nodes?

Example 2: Binary trees

B, the class of binary trees

Expecting B = • + • x SEQo,2(B)? Stay tuned

Theorem. If a simple variety of trees F = Z [ x or ★] SECMF)

Q. How many unary-binary trees with N nodes?

degrees of all nodes 0, 1, or 2

M, the class of all

M(z) = z( 1 + M(z) + M(z) 2 )

Theorem. If a simple variety of trees F = Z [ x or ★] SECMF)

Example 4: 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

A. N N ~ ] . (See EGF lecture.)

Example 4: Cayley trees (exact, from EGF lecture)

Construction C = Z*(SET(C)) <—

"a tree is a root connected to a set of trees"

EGF equation C(z) = ze C ^

by Lagrange inversion [z ]C(z) = [i/ ]( , )

with f(u) = u/e“ N u / e

Lagrange Inversion Theorem.

If a CF g(z ) = ^g„z n satisfies the equation z= f{g (z))

with f(0) = 0 and f ' (0) * 0 then gn = ~[u n

C N = N![z n ]C(z) =(n^) /

C, the class of all labelled rooted

Theorem. If a simple variety of trees F = Z [ x or ★] SECMF)

Aside: Stirling's formula via Cayley tree enumeration

Exact, via Lagrange inversion

Approximate, via singularity analysis

Example 4: Cayley trees

Theorem. If a simple variety of trees F = Z [ x or *] SEQ$(F)

Philippe Flajolet and

7. Applications of Singularity Analysis

• Simple varieties of trees

Philippe Flajolet and

7. Applications of Singularity Analysis

Simple varieties of trees

Transfer theorem for exp-log labelled set classes