Greedy algorithms
Download
1 / 51

Greedy Algorithms - PowerPoint PPT Presentation


  • 196 Views
  • Uploaded on

Greedy Algorithms. Greedy Algorithms. Compared to Dynamic Programming Both used in optimization problems More efficient Not always optimal Top-down instead of bottom up Make a locally optimal (quick) choice Still have optimal substructure No need to dive into sub-problems.

loader
I am the owner, or an agent authorized to act on behalf of the owner, of the copyrighted work described.
capcha
Download Presentation

PowerPoint Slideshow about ' Greedy Algorithms' - bailey


An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.


- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript
Greedy algorithms

Greedy Algorithms

Jeff Chastine


Greedy algorithms1
Greedy Algorithms

  • Compared to Dynamic Programming

    • Both used in optimization problems

    • More efficient

    • Not always optimal

    • Top-down instead of bottom up

  • Make a locally optimal (quick) choice

    • Still have optimal substructure

    • No need to dive into sub-problems

Jeff Chastine


Activity selection
Activity Selection

  • Given a set of activities , find the maximum # you can schedule

    • NOT optimal use of time of a room

    • One activity is active at one time

    • Activity has start time of and finish of

    • Order activities by finish time

isifi

1 2 3 4 5 6 7 8 9 10 11

1 3 0 5 3 5 6 8 8 2 12

4 5 6 7 8 9 10 11 12 13 14

Jeff Chastine


Using dynamic programming
Using Dynamic Programming

  • Defining the subproblem:

  • Set of activities that “fit” in this time chunk

  • Picking any activity creates two sub-problems

Jeff Chastine


Visualization
Visualization

Sij

aa

ac

ak

ax

az

ab

aw

ay

Jeff Chastine


Visualization1
Visualization

Skj

Sik

Sij

  • Optimal solution includes

aa

ac

ak

ax

az

ab

aw

ay

Jeff Chastine


Visualization2
Visualization

Skj

Sik

Sij

  • We know

  • How to choose activity ?

aa

ac

ak

ax

az

ab

aw

ay

Jeff Chastine


Visualization3
Visualization

Skj

Sik

Sij

  • Iterate through all (like Matrix Mults)!

aa

ac

ak

ax

az

ab

aw

ay

Jeff Chastine


Recursive solution
Recursive Solution

RECURSIVE-ACT-SELECT (s, f, i, n)

1 m ← i + 1

2 whilem ≤ n and sm < fi //find 1st activity

3 dom ← m + 1

4 ifm ≤ n

5 then return {am}

RECURSIVE-ACT-SELECT(s, f, m, n)

6 else return

Jeff Chastine


Code Trace

a0

0

1

2

3

4

5

6

7

8

9

10

11

12

13

14

Jeff Chastine


First Call

a0

a1

0

1

2

3

4

5

6

7

8

9

10

11

12

13

14

Jeff Chastine


Second Call

a2

a0

a1

0

1

2

3

4

5

6

7

8

9

10

11

12

13

14

Jeff Chastine


Second Call

a2

a0

a1

a3

0

1

2

3

4

5

6

7

8

9

10

11

12

13

14

Jeff Chastine


Second Call

a2

a0

a1

a4

a3

0

1

2

3

4

5

6

7

8

9

10

11

12

13

14

Jeff Chastine


Third Call

a0

a1

a4

0

1

2

3

4

5

6

7

8

9

10

11

12

13

14

Jeff Chastine


Third Call

a5

a0

a1

a4

0

1

2

3

4

5

6

7

8

9

10

11

12

13

14

Jeff Chastine


Third Call

a5

a0

a1

a4

a6

0

1

2

3

4

5

6

7

8

9

10

11

12

13

14

Jeff Chastine


Third Call

a5

a0

a1

a4

a6

a7

0

1

2

3

4

5

6

7

8

9

10

11

12

13

14

Jeff Chastine


Third Call

a5

a0

a1

a4

a8

a6

a7

0

1

2

3

4

5

6

7

8

9

10

11

12

13

14

Jeff Chastine


Fourth Call

a0

a1

a4

a8

0

1

2

3

4

5

6

7

8

9

10

11

12

13

14

Jeff Chastine


Fourth Call

a9

a0

a1

a4

a8

0

1

2

3

4

5

6

7

8

9

10

11

12

13

14

Jeff Chastine


Fourth Call

a9

a0

a1

a4

a8

a10

0

1

2

3

4

5

6

7

8

9

10

11

12

13

14

Jeff Chastine


Fourth Call

a9

a0

a1

a4

a8

a11

a10

0

1

2

3

4

5

6

7

8

9

10

11

12

13

14

Jeff Chastine


Result

a0

a1

a4

a8

a11

0

1

2

3

4

5

6

7

8

9

10

11

12

13

14

Jeff Chastine


Where greedy fails
Where Greedy Fails

  • Knapsack Problem

    • Steal items of different weight and value

    • Sack has weight limit

  • Goal: steal as much as possible

  • Two flavors of problem

    • 0-1 version (take all of something)

    • Fractional (can take only part)

Jeff Chastine


0 1 example
0-1 Example

50

item 3

30

item 2

20

item 1

10

$100

$60

$120

Jeff Chastine


Greedy go for the gold
Greedy – Go for the Gold!

50

item 3

30

item 2

20

item 1

10

$100

$60

$120

Jeff Chastine


Greedy go for the gold1
Greedy – Go for the Gold!

50

item 3

30

item 2

20

item 1

10

10

$100

$60

$120

Bad Idea…

Jeff Chastine


Greedy pick the next best
Greedy – Pick the next best

50

item 3

30

20

item 2

20

item 1

10

10

$100

$60

$120

Total = $160

Jeff Chastine


Greedy fail
Greedy Fail

50

30

item 3

30

item 2

20

item 1

10

10

$100

$60

$120

Try again! Total = $180

Jeff Chastine


Optimal for 0 1 problem
Optimal for 0-1 Problem

50

30

item 3

30

item 2

20

20

item 1

10

$100

$60

$120

Total = $220

Jeff Chastine


Fractional example
Fractional Example

50

item 3

30

item 2

20

item 1

10

$100

$60

$120

Jeff Chastine


Greedy go for the gold2
Greedy – Go for the Gold!

50

item 3

30

item 2

20

item 1

10

$100

$60

$120

Jeff Chastine


Greedy go for the gold3
Greedy – Go for the Gold!

50

item 3

30

item 2

20

item 1

10

10

$100

$60

$120

Jeff Chastine


Greedy get next best
Greedy – Get next best…

50

item 3

30

20

item 2

20

item 1

10

10

$100

$60

$120

Jeff Chastine


Greedy aha
Greedy – AHA!

Take only part

50

2030

item 3

30

20

item 2

20

item 1

10

10

$100

$60

$120

Jeff Chastine


Greedy go for the gold4
Greedy – Go for the Gold!

50

2030

$80

item 3

30

20

item 2

$100

20

item 1

10

10

$60

$100

$60

$120

Total = $240

Note: optimal for 0-1 was $220

Jeff Chastine


Why the greedy fail
Why the Greedy Fail?

50

50

30

item 3

30

20

item 2

20

item 1

10

10

10

$100

$60

$120

Notice the waste of space!

Jeff Chastine


Huffman codes
Huffman Codes

  • Objective: compress text data

  • Raw ASCII text file uses 8 bits/char (fixed)

    • 1000 characters = 8K bits

    • Book uses 6 characters for 3 bits/char

  • Trick

    • Analyze: not all characters are used!

    • Analyze: frequency of characters

    • Prefix codes: no binary code is prefix of another

Jeff Chastine


Analysis
Analysis

a b c d e f

Frequency (in thousands) 45 13 12 16 9 5

Fixed-length codeword 000 001 010 011 100 101

Variable-length codeword 0 101 100 111 1101 1100

110001001101 = face //12 bits

01011010 = abba // 8 bits

Jeff Chastine


How it works
How it works

  • Sort by least frequent

e:9

c:12

b:13

d:16

a:45

f:5

Jeff Chastine


How it works1
How it works

  • Group lowest two

e:9

c:12

b:13

d:16

a:45

f:5

Jeff Chastine


How it works2
How it works

  • Combine and re-sort

14

c:12

b:13

d:16

a:45

1

0

e:9

f:5

Watch how least frequent characters are pushed down the tree!

Jeff Chastine


How it works3
How it works

  • Group lowest two

14

c:12

b:13

d:16

a:45

1

0

e:9

f:5

Jeff Chastine


How it works4
How it works

  • Re-sort

14

25

d:16

a:45

1

1

0

0

e:9

f:5

c:12

b:13

Jeff Chastine


How it works5
How it works

14

25

d:16

a:45

1

1

0

0

e:9

f:5

c:12

b:13

Jeff Chastine


How it works6
How it works

25

30

a:45

0

1

1

0

14

c:12

b:13

d:16

1

0

e:9

f:5

Notice what’s happening to the first grouping we did

Jeff Chastine


How it works7
How it works

25

30

a:45

0

1

1

0

14

c:12

b:13

d:16

1

0

e:9

f:5

Jeff Chastine


How it works8
How it works

55

a:45

0

1

25

30

0

1

1

0

14

c:12

b:13

d:16

1

0

e:9

f:5

Jeff Chastine


How it works9
How it works

100

1

0

55

a:45

0

1

25

30

0

1

1

0

14

c:12

b:13

d:16

1

0

e:9

f:5

Jeff Chastine


Summary
Summary

  • Greedy algorithms have two properties

    • Greedy choice property: a globally optimal solution can be created by the greedy choice

    • Optimal substructure: the optimal solution is contains optimal solutions to subproblems

  • Greedy is top-down

  • Doesn’t always yield optimal solution

Jeff Chastine


ad