Midterm 1

Guide
Study

by George
 ③
Recurrence Tree
D
+ Substitution + Master's
Given Tcnl =
27 T(5) + 53 Recurrence Tree D

problem with original size 'U ·

is broken into 27 subproblems
·
Each Subproblem is size
·
5h"work done at each level of recursion
level
Depth Level work per
N

↓
O 543

i 2)

27 (5)
-
5(5)
=
= 27

729
·

:
5
5 =
543

543
umber of levels

Depth : Also known as

theheight of the tree The
.
recursion will
continue until the proble size becomes 1 (This is usually the base
case in Code like if IA 1 or
= if n 1) = =
,

: The
Work work/cost at each level is in

In this case : Tch) = Work at each level ·

Depth 5n (logn 1)
= ·
+

en
H is because for level O before recursion startsI
we are
accounting

TCH) simplifies to a O(logn)

You can also set it as a summation :
up
logn logn

5
5431
5430 logand
- =

=1
n is a constant here T =
1 n

using[ O(nPlog n)
=

,
· From Recurrence
,
we know Tcl =
Ochlogh) Substitution &

·
Use induction to prove -
1) Assume TCn)sCologn for some constant C

2) Substitute this hypothesis into our

You
Recurrence=n
can use

instead

Tcn) S27c(blog(5) + 543 Now Simplify

Tcn) S27 · Clog() + 5n =

ch
>
log() + 5n3 10g, 3 1=

-
Tcn) < Ch (logn - 10gs) 543 + =
Chlogn-unlogs + 53
=
chlogn cr + 53 -

As 'n grows ,
chlogn dominates so Tch) Sch logn

But let's solve forC

OChlogn)
TIIE chlogn - cr + 53 = Chlogn + n3(5 c) -

Ensure that Tcu) Schlogh

5 -
(So
,
which

No =3
*
will

>
only
<25
hold

due to base of
true

loggh
if :

↑
For C25 and
any n2lo the Recurrence Tcalschlogh holds.

Master's Theorem & I recommend using

this first ! Get Oc faster

T(n) =
27(b) + 5n3 nd us fen) add logn

a = 27

Calculated
b=

loga 3
 What about Tch) =
Tch-11 + 3h ?

Recurrence Tree &

depth Level Work

n -
i= 1
ou 3h U

n i I 3(n 1) 3(n -i)

n
= -

i= 1
Continues until the
base is reached 2 n 2 3(n 2) -
n

31
case
-

at 1 h times
,
d 3 (n-i)
Substitution &
↓ ↓
Prove Tcn) Ocu2)
3423 (n +
·
=
using Induction 0

Assume Con2
.
·

Tch) S
32 -
3
Substitute TCn)SCo(n-1 + 3n

342 (1 -

2) + E
Tch)3co(h2 2n 1) + 3 Ch2-2Ch#C
+whatic
- + = +

T (h) S (n2 + n(3 2C) -

+
L

dominates
On
Make sure n(3-2C) -3-2CSO
Because Cano exis
&ZSC Not so Tch = Ocas

Master's Theorem & Cannot USE Here !!!

Need to have a Recurrence in the form Tcul = aT(b) f(n) +

& =
# subproblems befactor of problem size reduction ful : work per call

·
Problem needs to be reduced LOGARITHMICALLY NOT LINEAR !
,
-
change in Depth
 For Pseudo code :

focus
·

on clarity &
simplicity
·
use clear a
descriptive variables

·
stick to universal constructs like if, else, for, while return
,

State
inputs and
outputs
·

·
Aim for readability basic programming concepts
,

* Show understanding of the

question explain solution approach
,

* Provide runtime analysis

If it helps , you can answer these

questions in
paragraph
format Show . us your understanding of the
question and concepts
to
get the solution .

some algorithms to know :

Majority of these HW
Merge Sort Merge problems
·
+ were

·
List all subsets of an
array list (Brute force)
·

finding smallest number in list A

of list
Finding sum elements

&
·
in

·
List all substrings given a
string
Binary Search · Divide &
Conquer ways to do these
·

MAYBE MaxSubarray if covered

·