25
1 Presented by: Yuchen Bian 4.16.2015 RWC: lustering based on Multiple Random Walks Cha

1 Presented by: Yuchen Bian 4.16.2015 MRWC: Clustering based on Multiple Random Walks Chain

Embed Size (px)

Citation preview

Page 1: 1 Presented by: Yuchen Bian 4.16.2015 MRWC: Clustering based on Multiple Random Walks Chain

1

Presented by: Yuchen Bian

4.16.2015

MRWC: Clustering based on Multiple Random Walks Chain

Page 2: 1 Presented by: Yuchen Bian 4.16.2015 MRWC: Clustering based on Multiple Random Walks Chain

2

1. Introduction and Motivation----Background

2. Multiple Random Walks Chain (MRWC)----Intuition----Definitions

3. Experiments

4. Conclusion

5. Future Work

Content

Page 3: 1 Presented by: Yuchen Bian 4.16.2015 MRWC: Clustering based on Multiple Random Walks Chain

3

1. Introduction and Motivation----Background

2. Multiple Random Walks Chain (MRWC)----Intuition----Definitions

3. Experiments

4. Conclusion

5. Future Work

Content

Page 4: 1 Presented by: Yuchen Bian 4.16.2015 MRWC: Clustering based on Multiple Random Walks Chain

4

1. Introduction and Motivation

Random Walk Model:

a

b

c

1

1/2

1/21

t=0

a

b

c

1

1/2

1/21

t=1

a

b

c

1

1/2

1/21

t=2

a

b

c

1

1/2

1/21

t=3

Page 5: 1 Presented by: Yuchen Bian 4.16.2015 MRWC: Clustering based on Multiple Random Walks Chain

5

5

xt+1(i) = ∑j(Probability of being at node j)*Pr(j->i) =∑jxt(j)*P(j,i)

xt+1 = PTxt

Long time after…

xt+1 ≈ xt

xt = PTxt

Converge to a stationary distribution π no matter what the initial distribution is.

For each πi

πi=d(i)/2m

1. Introduction and Motivation

Random Walk Model:

Page 6: 1 Presented by: Yuchen Bian 4.16.2015 MRWC: Clustering based on Multiple Random Walks Chain

6

1. Introduction and Motivation

Random Walk Model:

πi=d(i)/2m

Query node: 8

Page 7: 1 Presented by: Yuchen Bian 4.16.2015 MRWC: Clustering based on Multiple Random Walks Chain

7

7

xt = PTxt

ei is a vector in which only the i-th (query node) element is 1, otherwise 0.------Restart

c 0≤c<1

1. Introduction and Motivation

Random Walk with Restart Model:

xt = (1-c)PTxt+cei

Page 8: 1 Presented by: Yuchen Bian 4.16.2015 MRWC: Clustering based on Multiple Random Walks Chain

8

1. Introduction and Motivation

Query node bias: sharp peak

Query node: 8

Random Walk with Restart Model:

Page 9: 1 Presented by: Yuchen Bian 4.16.2015 MRWC: Clustering based on Multiple Random Walks Chain

9

1. Introduction and Motivation

For large graph, convergence needs more time. Query node: 8

Local clustering:

Find cluster before convergence, even the RW will not reach some nodes.

In fact, a RW might be restricted in the cluster with high probability, HOWEVER, it is also hard to travel back if RW pass through boundary

Targets: restricted in the cluster which contains the query nodes.

What if the query node(s) send out a series of RWs, not a single RW, hopefully, this RWs group is harder than single RW to travel through boundary.

Page 10: 1 Presented by: Yuchen Bian 4.16.2015 MRWC: Clustering based on Multiple Random Walks Chain

10

1. Introduction and Motivation----Background

2. Multiple Random Walks Chain (MRWC)----Intuition----Definitions

3. Experiments

4. Conclusion

5. Future Work

Content

Page 11: 1 Presented by: Yuchen Bian 4.16.2015 MRWC: Clustering based on Multiple Random Walks Chain

11

Intuition:

2. Multiple Random Walks Chain (MRWC)

From each query node, send a series of RWs to explore the graph, all RWs walk one by one,

but the next vertex the current RW will explore is not only follow its own “thought” but also decided by other RWs.

Then all RWs constructs a RWs group and this group is harder than a single RW to travel through the boundary.

Page 12: 1 Presented by: Yuchen Bian 4.16.2015 MRWC: Clustering based on Multiple Random Walks Chain

12

Definitions:

2. Multiple Random Walks Chain (MRWC)

Definition 1: [Multi-Random Walks Chain (MRWC)]

A MRWC is that for each query vertex, from time point 0, sending k random walks at following time point 1, 2,..., k. At time point , assume that the k random walks stand at vertices .

At each time point, there is only one random walk which follows the order to and recursively searching through the graph, and at the same time, other random walks have effects on the next vertex the current random walk will go to.

Page 13: 1 Presented by: Yuchen Bian 4.16.2015 MRWC: Clustering based on Multiple Random Walks Chain

13

Definitions:

2. Multiple Random Walks Chain (MRWC)

Page 14: 1 Presented by: Yuchen Bian 4.16.2015 MRWC: Clustering based on Multiple Random Walks Chain

14

Definitions:

2. Multiple Random Walks Chain (MRWC)

Page 15: 1 Presented by: Yuchen Bian 4.16.2015 MRWC: Clustering based on Multiple Random Walks Chain

15

1. Introduction and Motivation----Background

2. Multiple Random Walks Chain (MRWC)----Intuition----Definitions

3. Experiments

4. Conclusion

5. Future Work

Content

Page 16: 1 Presented by: Yuchen Bian 4.16.2015 MRWC: Clustering based on Multiple Random Walks Chain

16

3. Experiments

Computation and Egs:

A3, P3

1

2

3

4

5

6

7

8

9

11 16

1312

17

15

20

1819

1410

Naïve Method:Iteratively computation

Page 17: 1 Presented by: Yuchen Bian 4.16.2015 MRWC: Clustering based on Multiple Random Walks Chain

17

17

1

2

34

5

6

7

8 910

11

12

13

141516

17

20 21

32

33

1819

26 27

2425

2223

3031

3435

2829

38

39

3637

4041

4243

44

45

48

49

46

47

50

51

5253 55

54

5657

58

59

60

A2, P2

Page 18: 1 Presented by: Yuchen Bian 4.16.2015 MRWC: Clustering based on Multiple Random Walks Chain

18

3. Experiments

Fig 1. Basic RW Fig 2. RWR

Fig 3. MRWC (k=2) Fig 4. MRWC (k=3)

Page 19: 1 Presented by: Yuchen Bian 4.16.2015 MRWC: Clustering based on Multiple Random Walks Chain

19

3. Experiments

Not converge! Why?Dynamically choosing the other RWs’ position with largest probability

MRWC (k=2)RWs’ position for each iterationW1-B*, W2-Rs

Page 20: 1 Presented by: Yuchen Bian 4.16.2015 MRWC: Clustering based on Multiple Random Walks Chain

20

20

3. Experiments

Fig 1. RWR

Fig 3. MRWC (k=3)

Fig 2. MRWC (k=2)

RWs’ position for each iteration (k=2)

Page 21: 1 Presented by: Yuchen Bian 4.16.2015 MRWC: Clustering based on Multiple Random Walks Chain

21

1. Introduction and Motivation----Background

2. Multiple Random Walks Chain (MRWC)----Intuition----Definitions

3. Experiments

4. Conclusion

5. Future Work

Content

Page 22: 1 Presented by: Yuchen Bian 4.16.2015 MRWC: Clustering based on Multiple Random Walks Chain

22

4. Conclusion

• Motivation: restrict into the target cluster• Advantages:

• Increase the number of features• Sharpen the boundary: harder to pass through than single RW• Group activity not single activity (sharp peak)

• Disadvantages:• Convergence issue• Naïve method

Evaluation to MRWC:

Page 23: 1 Presented by: Yuchen Bian 4.16.2015 MRWC: Clustering based on Multiple Random Walks Chain

23

1. Introduction and Motivation----Background

2. Multiple Random Walks Chain (MRWC)----Intuition----Definitions

3. Experiments

4. Conclusion

5. Future Work

Content

Page 24: 1 Presented by: Yuchen Bian 4.16.2015 MRWC: Clustering based on Multiple Random Walks Chain

24

5. Future Work

• Model: • formal and general model

• Mathematical Analysis:• Convergence?• How to sharpen the boundary?

• Algorithm:• Efficient computation or approximation• Compare with other methods

Page 25: 1 Presented by: Yuchen Bian 4.16.2015 MRWC: Clustering based on Multiple Random Walks Chain

25

Yuchen Bian

Thank you! Q & A