Advanced data

Structures and

algorithms
Notes
 Mus
>
Splay Notes :
-

Partitioning
>
-

stays at
different levels of partitioning

h or
hypergraph
-usually with vertex
weight/size
>
usually weighted edges
- .

straintspartitioning
s

(K-way partitioning)
>
-

MaximumCapacity of
each
partitiono
maximum allowable
difference between
partition
.

objective odes to
partitions subject
.
to constraints such that
there

cutsize minimized
is

Fractabil,ity p-complete .

Kernighan Lin (KL)

algorithm
>
-
-

calculation.
Gain
>
-
Fontusios -

Matthesus
algorithm
 >
-

Weighted event
scheduling
> Matrix chain
Multiplication
-

Booth
-
& Wallace
multiplies
>
-
storisius matrix
multiplication ?
> show that no
of paranthesis grows
at
(4/ ?
-

Longest subsequence
>
-

common

The
knapsack problem
>
-

O knapsack problem

Randomized Las
Vegas (Always
correct solution)
>
Algorithm >
gives
-

Montecarlo incorrect solution

>
(May give
> Randomized
Quicksort
↓I
onesided Trosidas
e

-min-cut
algorithm
-
Kargar algorithm partition
>
-

Karnighan Lin algorithm ↓ Y

>
-

Fedusia Matthesus Balanced Unbalanced

Simulated
Annealing
-

>
-

Simulated Evolution
>
probability computing Michael M
.
-

Markov's
>
Inequality
.
-

Chebyshev's Inequality
>
-

Chernoff's bound
.

>
-

Weighted vertex cover

problem (WVC)
>
Travelling salesman
problem (TSP)
-

> Hamiltonian cycle problem (HCP)

-

-> Eulerian
graph
The
K-clustering problem
>
-

# knapsack problem

- circuit SAT
(Questions willcome
Sop POS
format

>
polynomial Reducibility
-

NP-complete problems
>
-

>
-
3- SAT

>
3-colorability problem
-

>
- The
clique problem
>Vertex cover
problem
-
 shortest path with Hedge
b for v
path s-x
① => every
,

② b Bellman-Ford Algorithm O(mn)

③ b

#min
C -
+ 1 = 33

, = 1

13 11 10

-
②
---
⑤ b

⑦ C

111/
⑧ a Randomized Select
Algorithm Oct.
worst A
but case
- 0(n) .

⑨ (9) True

every bipartite
(b) False becoz Not
graph
is tree But tree is
every
.

bipartite graph .

False An undirected have

( graph can

multile path possible .

 (d) True As M5T have
unique
edge to connect .
(No
cycles .

false plogs
es as => n)
logn .

Oth
complexity
So , is Time

False
(f) it solved .
DP
as is
using 2-D ,

True
(g)
as minimum
M5T
edge is the
part
of
(h) True

i True

false Cas AXL & Retre is

O(legn))
(j) .

⑩
 Common
LCS7
Longest Subsequence
.

Time
=>
Complexity of LCS
for Dynamic method : -MXn

if

S
k or t
[FCs (KIst
= 0 = 0

Les (1 A]
,
=
+]
if 10 , 30, A

max[LC5 (k-1 , AJ ,
LCS (k -1)
,

if 170 , 130 , Un = ya

t
↓

max
subsequent
length
.

A
Subsequence :
=> Time
complexity of LCS
for recursive method :

nxzm or mx2"

E
1 + (C8(i 1) if i
j j
- 1 -

=
,

mar (LCS (i -1 j)
LCS ,

if i
j
= ,

225(i
, j 1))
-

123456
if i = 0
,

j
.
= 0

Q X =
BDGABA
Y = ABC BDA
⑭ 3
13
5)
[S2
LCs (6 6) 1+ LC5 (5 - 25
,
=
, , 2) = man
3

(LCS
LCs (5 5) ,
= mar (4, 53
, 2

I
LC5 (5 4) 3 2C5 (2 3)
GLC5(1 3) 1
,
,
=
man ,
2

[25(3
5)2 LC5 (2, 2) I
LOS (4 5) ,
= may ,
↑
S
LC5 (4 , 4)2 (C5 (4 , 1) = 1 + LC5 (3 07
,

3 # ⑧
LCs (5 , 4) =
1 + LCS (4 3) 2
,
-
LC5(1 , 2) = 1 + LC5(0, 1) -

z O
LCs (3 5) ,
=
man

((C512
LC5 (3
, 5)

4)2
,
2 LC5(2 , 1) =
man
&LC5
LC5 (2, 0)
(1 1178 ,

o
,

z O
(C5 (44)
(105(3 4)2 LC5(3 , 1)
man Las
=
,
= man ,

LCS (4, 31 2
I

(2C5 (2C510
225 (3 ,) =
man (3 3)2 , LC5(1 / 3) = man , 3)0
LC5 (2 , 4) I LC5 (1 , 2) 1

z
3)
E (LCSto
LC5 (4 , 3) =
mas LCS (3 3) .
2
LC5 (1 , 1) = man ,

28(10)
2C5(4, 2) 1

Las (24,5) = 1 + 1C5 (1 , 4)1

(333) 2)
substring
V
LC5 = 1 + Los (2 , mak

[15(1
*
431
length
=4
LCs (2 , 4) = mar , .
LC5 (2 , 3) 1

205" (4 2)
(LC5(b 2)
Substringe BCBA
=
, man ,

(C8(4 1) ,
I

205"<1 , 4) = 1 + LCS (0, 3)

O

I p
225 (2 2) man

&Les c,0
=
,
 chain
=> Given a
of matrices (A ,, Az -
--Ax]
where A has dimensions Pix XPi ,

Fully
parenthesize the
product A2---An Al ,

in a
way that minimizes the number
of multiplications
scalar . Give the recursive
solution
for minimizing the cost of
parenthesizing the Matrix Ai Ai 1---Aj and .
+

compute
Al XA2 x A3 X A4
wher ,
Al : 2X4 A2 : 4x12 A3 : 12X4 A4 : 4x10

- The
following recursive solution is

Emin
i
[i
g if =
j
e
j]
=
,

[c(i , ) +
e (k ,
j
+

Pipe
AiXA2 X Az XA4

-
2x44X12 12x44X10

↑
=
min [0 + 0 + 2x4x123 96 =

min(e (2 p xpzXP33

t
c (3 37
↓ (2 3)
,
=
, 2) +
, +
,

=
min[0 + 0 +
4x12X43 = 192 .

C (3,4) = m [ e(3 3) , + c(4 4) ,

+ P2XPzXP+ ]
= min 20 + 0 + 12x4X103 = 480

mintboxbo- (1 , 3)
,A
= = 0 + 192 + 2x4X4 = 224
c (1 , 23 + c (3 3)
, + prxb2xP3
1 < 3 -1
C(
2k4
,) =

meet
E c (2 3) , + c(4, 4) + P
, XPzXP4
0 = 480 + 48 = 96

192 + 0 4X4X10
=
+ = 192 + 160 =
352

min
S F
0 + 352 + 2 x4 X 1 0 = 352 + 80 =
432

e ,) =
9) + 480 + 2x12x10 = 96 + 480 + 240 = 816

1 < k <4 c (1 , 3) + c (4 4]
,
+ Po XPzX Pr
2 + 00 = 22

Parenthesization .

S 3

(2A , xAcxAs) X A

An
Que Karger min-cut algorithm : -

o
>
To collapse the edge Cr v) create node
-

we a new
, ,

Ey replace any edge of the form

,

En wa or
, Ev ws ,
with a new
edgecur ,
we
and
Mand
v
then delete the
original virtices .

Min cut
>
Karger :-
-

1: Run
return
the
the
guessing
smallest
algo many . times and

guess
.

Only wayis for Karger algo return the

2: to
wrong
answer
if Karger-min-cut algo fails N times

in a row
.

Karger Min Cut (G) :

mink

i
for 1 to N .:
=

i = Guess Min Cut (4)

mink
if 1x1 < .

mink = 1x1

mina = X .

return mink .
 Applying min-cut-algo to
given ques :

rapsing↓
( , 4) Yater pai

- collapsing 20

collapsing Ga

#D a re
final graph
after applying
the
algorithm
.

Qubi-

F y2
=> expression for
firstly deriving its Boolean the above circuit :

Yo yay - %
=

Now * 5
y
=
,
yy
=
y5xys
-

y
= x , nx4 , y2
=
T

yz
=
nxn T yq
=
(4 144)
,
*
42

T
-

②
y5
=

77 2
x(uznx)
=

>5]
Now , (((4 , 2x4Xx2) /Ev(Uza ** )) <

ye
=

Now in CNF : -Conjuctive Normal Form "

for
yo
to be .

using the following identify :-

(an b) y c =
(avc)n(byc) .

%o
=
((k xx2) n (44YX2)]
, /zvU3) NLT2 vn] nT5

the circuit
Now for to be satisfiable following conditions
must be hold true :

I Tre
(k2VU , ) >
-

must be true

(x2 XX4) >

-

True True
(V *) >
-
True
 then for (x2X4 , ) >
-

true

considering/
true l = then , I, can be
anything

(O2Vx) >
-

true since we let xe-true

.

(v) >
-

for this to be true :-

-False >

is >
-
true
so,
Iuse
(2 YUz) >
-

for this to be true

y has to be true

#-
true .

Hence , 41 >
-

True/false
42 - True

3 - True
44 >
-
False
45 - False

for all these

input our circuit will be satisfied
.

Ques +
WSCP
by pricing method : -Given
diagram
is :

w
b
w(S1) =
So the WSCP that :-
,
says
i/p = a set X =
Ex , x2 ,
- - - - - -
kn] of n elements .

a
family f =

[A ,, 12 , As ---- Sny of m subsets

of u and w(f) =
weights of each

subset If .

w(f) > 0 .

Pre-condition =f covers a i .

.
e

o p = a subset eef! that covers be Uses

Goal :- Minimize the weight of set cover C:

w(e)= was

Now
coming
to our
question : .
-

w(s)
⑭
-

·
&

③>
- 12

>
-
15
⑭
⑮ > 10
⑳
-
Now the optimal solution :

Copt
=
[S2 , Az , 15]

W(Coptl = 9 + 12 + 10 = 31 .

Now approx WSCP

by ispricing method :-

the
algorithm as
follows :

E
N = ( *
uncovered elements)
c d
= ( set cover)

while U :

select sef that minimizes pricp .

=it
a

noo
s demn

u = v -
S

c = c S

return C

#
Running the above
algorithm ,
ituati
#
on

p
=
2475 , >
-

% = 2 -
X-

Ya
v
2

2X3 > 9/4 2 25

-

= .

2 *4 Ss
> 12/3 4
-

X5
Sk > 15/2 = 7 5
X6
- .

X7 Sq
Xf
>
-
1 % =
3 .
33

itration o
d
& p =

u
2 K 1, a
,

2
9/2 = 4 5 .

12/1 = 12

15/1 = 15
4 5 .

Us
4 .

5 s
1 %
/r = 5

RT
Rf

iteration
Erd
ice x
p =
,
Si a

2
R2
52 es

5i as
Sy

Sk
12/1 =
12

4 5 1/1 =
15
XS
.

x7 S5 10/1 = 10
E
1048
-
 4th iteration
-

Price : -

24 ,
Si
D
2xz S2 s

e-in
s
53 e

In -

12x7
10 Up

EPi(ki) w(l) 4
2 + 2 + 54-
Now 2 + 2
= = + .

4
. 5 + 12+ 10

(c)
39 =

algorithm
using
Etopt = 31
.

=> Que: Karnigham Lin (KL)

algorithm : E
determining the Minimum Cut :-

-Given graph :-

a
Step
I
-p
I
find the pair of verti where
cut size decreases :

Let us take :
vertise pair (H , ED

se
Now cut =
2 .

#Nowagain taking pai

a verter s

⑭
Now cut =
1

So 2-1
gain
=

= 1 .

So ,
here also : -cut has been decreased
Hence also accepted
, this
swapping is .

Now if we
swap more
pair of vertices
,
will
definitely the
then cut increase
.

This
I is the minimum that we can

get .
proteo
verters cover
for
-

above =
⑦2

[B(4) ,
C(2) ,
e(2)}.
↓
*
total weight = 8 .= c

c =
0 .

Now ,
approximation algo, verte cover : -

1st
edge > (A C)
-

e =
qU[A CY ,

⑦-----

·
2nd
edge ,
c =
&A CSU (B D)
,
,

C =
[A , C , B D3 > total
weight
-

>
-

3 + 2 + 4# 4
·

> 13 = C
⑰
-

----------
des
 Here
,
co =
13
*
C = 8

s
>
pans
-

C cost of solution

of optimal
*
c >
-
cost solution
For minimization : -

i
for

maximizati
I
STATION
·

·
S1 ,5

LINE-I

e Exit
↑
Entry
- ⑫
⑫
5
X
5
654

S 52 ,

STATION
2 S2 , 3

STATION
52 , 4

STATION
S2 ,

STATION
5

if j
detai
=
2
Fli] =

[F/j-1)
min +
Rej F2/j-c] + +
, ,
+
is a] if j22
 E/j) =Let Feli-t] + + y
,
-

1
+
azifj22
For j = 1

# 11] 8
= e + au =
2+ 6
,
=

5+ 8 13
F(1] e + 22
=
= =

For j = 2

F(2) = min GFC1] + 42

,
F2(1] + te + &123
,

= min
[0 + 8 ,
13 + 2 + 8]
= 16

GE [11 + &22 , F(1] +A, c am2)

+
F2(2) = min

= min
[15 + 5 0 + + 6),
2

= 16