26
CMU 15-251 Online Algorithms Teachers: Anil Ada Ariel Procaccia (this time)

PowerPoint Presentationaada/courses/15251f15/www/slides/lec17.pdfSki rental ≥8. 1. 𝐺𝐼= t⋅ 𝑇𝐼 2. 𝐺𝐼= u⋅ 𝑇(𝐼) 3. 𝐺𝐼= 𝐵 2 ⋅ 𝑇(𝐼) 4. 𝐺𝐼=

  • Upload
    lyminh

  • View
    222

  • Download
    0

Embed Size (px)

Citation preview

CMU 15-251

Online Algorithms

Teachers:

Anil Ada

Ariel Procaccia (this time)

Ski rental

𝐵

• 𝐵 = 5

2

Ski rental

• 𝐵

3

⇒ 10

⇒ 4

Ski rental

𝐵 ≥ 8.𝐵

1. 𝐴𝐿𝐺 𝐼 = 2 ⋅ 𝑂𝑃𝑇 𝐼

2. 𝐴𝐿𝐺 𝐼 = 3 ⋅ 𝑂𝑃𝑇(𝐼)

3. 𝐴𝐿𝐺 𝐼 =𝐵

2⋅ 𝑂𝑃𝑇(𝐼)

4. 𝐴𝐿𝐺 𝐼 = 𝐵 ⋅ 𝑂𝑃𝑇(𝐼)

4

Competitive ratio

• c > 1 𝐴𝐿𝐺𝑐 𝐼𝐴𝐿𝐺 𝐼 ≤ 𝑐 ⋅ 𝑂𝑃𝑇(𝐼)

• 𝑐 < 1 𝐴𝐿𝐺𝑐 𝐼𝐴𝐿𝐺 𝐼 ≥ 𝑐 ⋅ 𝑂𝑃𝑇(𝐼)

𝐴𝐿𝐺 𝑂𝑃𝑇(𝐼)

5

Ski rental, revisited

• 𝐵 − 12𝐵−1

𝐵

6

Ski rental, revisited

𝛼 𝛼 <2𝐵−1

𝐵

o 𝐾𝐾 + 1

o 𝐾 + 2

o 𝐾 ≥ 𝐵:𝑂𝑃𝑇 𝐼 = 𝐵, 𝐴𝐿𝐺 𝐼 = 𝐾 + 𝐵 ≥ 2𝐵

o 𝐾 ≤ 𝐵 − 2: 𝑂𝑃𝑇 𝐼 = 𝐾 + 1,𝐴𝐿𝐺 𝐼 = 𝐾 + 𝐵 ≥ 2𝐾 + 2 ∎

7

Pancakes, revisited

8

𝐵2𝐵−1

𝐵

Ski rental, revisited

9

Paging

• 𝑁 𝑘

10

Paging

11

Paging

12

Paging

1,… , 𝑘)

13

Example: LIFO

14

Paging

• 𝛼𝛼

1. 𝛼 = 2

2. 𝛼 = 𝑘

3. 𝛼 = 𝑁

4. 𝛼 = ∞

15

Paging

• 𝛼𝛼

1. 𝛼 = 2

2. 𝛼 = 𝑘

3. 𝛼 = 𝑁

4. 𝛼 = ∞

16

Paging

• 𝑘

o

𝑘

o 𝑘 = 3:

17

Paging

• 𝑘

o 𝑚 = 𝑝𝑗𝑖 𝑗

𝑖

o 𝑝1𝑖 , … , 𝑝𝑘

𝑖 , 𝑝1𝑖+1

o 𝑝2𝑖 , … , 𝑝𝑘

𝑖

𝑝1𝑖+1

⇒𝑂𝑃𝑇 ≥ 𝑚

18

Paging

• 𝑘

o

o 𝐴𝐿𝐺 ≤ 𝑘𝑚 ∎

19

Paging

• 𝑘

• ∎

•𝛼 𝛼 < 𝑘

20

Paging

o

{1, … , 𝑘 + 1} ⇒

21

Paging

o 𝑘∎

22

List update

• 𝑛

23

List update

24

List update

• 𝛼𝛼

25

Summary

o

o

o

o

26