CS246: Mining Massive Datasets Jure Leskovec, Stanford University

Transcription

1 CS246: Mining Massive Datasets Jure Leskovec, Stanford University

2 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 2 Web pages are not equally important vs. We already know: Since there is large diversity in the connectivity of the webgraph we can rank the pages by the link structure vs.

3 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 3 We will cover the following Link Analysis approaches to computing importances of nodes in a graph: Page Rank Hubs and Authorities (HITS) Topic-Specific (Personalized) Page Rank Web Spam Detection Algorithms

4 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 4 Idea: Links as votes Page is more important if it has more links In-coming links? Out-going links? Think of in-links as votes: has 23,400 inlinks has 1 inlink Are all in-links are equal? Links from important pages count more Recursive question!

5 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 5 Each link s vote is proportional to the importance of its source page If page p with importance x has n out-links, each link gets x/n votes Page p s own importance is the sum of the votes on its in-links p

6 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 6 A vote from an important page is worth more A page is important if it is pointed to by other important pages Define a rank r j for node j r j = i j r i d out (i) a/2 a The web in 1839 a/2 y/2 y y/2 m Flow equations: r y = r y /2 + r a /2 r a = r y /2 + r m r m = r a /2 m

7 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 7 3 equations, 3 unknowns, no constants No unique solution All solutions equivalent modulo scale factor Additional constraint forces uniqueness y + a + m = 1 y = 2/5, a = 2/5, m = 1/5 Flow equations: r y = r y /2 + r a /2 r a = r y /2 + r m r m = r a /2 Gaussian elimination method works for small examples, but we need a better method for large web-size graphs

8 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 8 Stochastic adjacency matrix M Let page j has d j out-links If j i, then M ij = 1/d j else M ij = 0 M is a column stochastic matrix Columns sum to 1 Rank vector r: vector with an entry per page r i is the importance score of page i i r i = 1 The flow equations can be written r = M r

9 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 9 Suppose page j links to 3 pages, including i j i 1/3 = i M r r

10 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 10 The flow equations can be written r = M r So the rank vector is an eigenvector of the stochastic web matrix In fact, its first or principal eigenvector, with corresponding eigenvalue 1

11 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 11 a y m y a m y ½ ½ 0 a ½ 0 1 m 0 ½ 0 r = Mr r y = r y /2 + r a /2 r a = r y /2 + r m r m = r a /2 y ½ ½ 0 y a = ½ 0 1 a m 0 ½ 0 m

12 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 12 Given a web graph with n nodes, where the nodes are pages and edges are hyperlinks Power iteration: a simple iterative scheme Suppose there are N web pages Initialize: r (0) = [1/N,.,1/N] T Iterate: r (t+1) = M r (t) Stop when r (t+1) r (t) 1 < ε x 1 = 1 i N x i is the L1 norm Can use any other vector norm e.g., Euclidean r ( t+ 1) j = i j ( t) i r d d i. out-degree of node i i

13 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 13 Power Iteration: Set r j = 1/N r j = i j r i d i And iterate r i = j M ij r j Example: r y 1/3 1/3 5/12 9/24 6/15 r a = 1/3 3/6 1/3 11/24 6/15 r m 1/3 1/6 3/12 1/6 3/15 Iteration 0, 1, 2, a y m y a m y ½ ½ 0 a ½ 0 1 m 0 ½ 0 r y = r y /2 + r a /2 r a = r y /2 + r m r m = r a /2

14 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 14 Imagine a random web surfer: At any time t, surfer is on some page u At time t+1, the surfer follows an out-link from u uniformly at random Ends up on some page v linked from u Process repeats indefinitely Let: p(t) vector whose i th coordinate is the prob. that the surfer is at page i at time t p(t) is a probability distribution over pages r j i 1 i 2 i 3 = j i j d out ri (i)

15 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 15 Where is the surfer at time t+1? Follows a link uniformly at random p(t+1) = M p(t) Suppose the random walk reaches a state p(t+1) = M p(t) = p(t) then p(t) is stationary distribution of a random walk Our rank vector r satisfies r = M r So, it is a stationary distribution for the random walk i 1 i 2 i 3 j p( t + 1) = M p( t)

16 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 16 r ( t+ 1) j = i j r ( t) i d i or equivalently r = Mr Does this converge? Does it converge to what we want? Are results reasonable?

17 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 17 a b Example: r a = r b Iteration 0, 1, 2, r ( t+ 1) j = i j r ( t) i d i

18 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 18 a b Example: r a = r b r ( t+ 1) j = i j r ( t) i d i Iteration 0, 1, 2,

19 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 19 2 problems: Some pages are dead ends (have no out-links) Such pages cause importance to leak out Spider traps (all out links are within the group) Eventually spider traps absorb all importance

20 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 20 Power Iteration: Set r j = 1 r j = i j r i d i And iterate a y m y a m y ½ ½ 0 a ½ 0 0 m 0 ½ 1 r y = r y /2 + r a /2 r a = r y /2 Example: r y 1/3 2/6 3/12 5/24 0 r a = 1/3 1/6 2/12 3/24 0 r m 1/3 3/6 7/12 16/24 1 Iteration 0, 1, 2, r m = r a /2 + r m

21 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 21 The Google solution for spider traps: At each time step, the random surfer has two options: With probability β, follow a link at random With probability 1-β, jump to some page uniformly at random Common values for β are in the range 0.8 to 0.9 Surfer will teleport out of spider trap within a few time steps y y a m a m

22 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 22 Power Iteration: Set r j = 1 r j = i j r i d i And iterate Example: r y 1/3 2/6 3/12 5/24 0 r a = 1/3 1/6 2/12 3/24 0 r m 1/3 1/6 1/12 2/24 0 Iteration 0, 1, 2, a y m y a m y ½ ½ 0 a ½ 0 0 m 0 ½ 0 r y = r y /2 + r a /2 r a = r y /2 r m = r a /2

23 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 23 Teleports: Follow random teleport links with probability 1.0 from dead-ends Adjust matrix accordingly y y a m a m y a m y ½ ½ 0 a ½ 0 0 m 0 ½ 0 y a m y ½ ½ ⅓ a ½ 0 ⅓ m 0 ½ ⅓

24 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 24 r + = Mr ( t 1) ( t ) Markov Chains Set of states X Transition matrix P where P ij = P(X t =i X t-1 =j) π specifying the probability of being at each state x X Goal is to find π such that π = P π

25 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 25 Theory of Markov chains Fact: For any start vector, the power method applied to a Markov transition matrix P will converge to a unique positive stationary vector as long as P is stochastic, irreducible and aperiodic.

26 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 26 Stochastic: Every column sums to 1 A possible solution: Add green links S = M + a T ) a y m y a m y ½ ½ 1/3 a ½ 0 1/3 m 0 ½ 1/3 1 ( 1 n a i =1 if node i has out deg 0, =0 else 1 vector of all 1s r y = r y /2 + r a /2 + r m /3 r a = r y /2+ r m /3 r m = r a /2 + r m /3

27 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 27 A chain is periodic if there exists k > 1 such that the interval between two visits to some state s is always a multiple of k. A possible solution: Add green links y a m

28 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 28 From any state, there is a non-zero probability of going from any one state to any another A possible solution: Add green links y a m

29 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 29 Google s solution that does it all: Makes M stochastic, aperiodic, irreducible At each step, random surfer has two options: With probability 1-β, follow a link at random With probability β, jump to some random page PageRank equation [Brin-Page, 98] r j = (1 β) r i i j d i Assuming we follow random teleport links with probability 1.0 from dead-ends + β 1 n d i out-degree of node i

30 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 30 PageRank equation [Brin-Page, 98] r j = (1 β) r i d i i j The Google Matrix A: + β 1 n A = 1 β S + β 1 1 1T n G is stochastic, aperiodic and irreducible, so r (t+1) = A r (t) What is β? In practice β =0.15 (make 5 steps and jump)

31 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 31 y 0.8 ½+0.2 ⅓ S 1/2 1/ / /2 1 1/n 1 1 T 1/3 1/3 1/3 1/3 1/3 1/3 1/3 1/3 1/3 a m ⅓ y 7/15 7/15 1/15 a 7/15 1/15 1/15 m 1/15 7/15 13/15 A y a = m 1/3 1/3 1/ /33 5/33 21/33

32 Suppose there are N pages Consider a page j, with set of out-links O(j) We have M ij = 1/ O(j) when j i and M ij = 0 otherwise The random teleport is equivalent to Adding a teleport link from j to every other page with probability (1-β)/N Reducing the probability of following each out-link from 1/ O(j) to β/ O(j) Equivalent: Tax each page a fraction (1-β) of its score and redistribute evenly 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 32

33 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 33 Construct the N x N matrix A as follows A ij = β M ij + (1-β)/N Verify that A is a stochastic matrix The page rank vector r is the principal eigenvector of this matrix A satisfying r = A r Equivalently, r is the stationary distribution of the random walk with teleports

34 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 34 Key step is matrix-vector multiplication r new = A r old Easy if we have enough main memory to hold A, r old, r new Say N = 1 billion pages We need 4 bytes for each entry (say) 2 billion entries for vectors, approx 8GB Matrix A has N 2 entries is a large number! A = A = β M + (1-β) [1/N] NxN ½ ½ 0 ½ ½ = 1/3 1/3 1/3 1/3 1/3 1/3 1/3 1/3 1/3 7/15 7/15 1/15 7/15 1/15 1/15 1/15 7/15 13/15

35 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 35 r = A r, where A ii = β M ii + 1 β N r i = j=1 A ii r j r i = N β M ii + 1 β j=1 r N j = N β M ii r j + 1 β j=1 N r N j=1 j = N β M ii r j + 1 β j=1, since r N j = 1 So, r = β M r + 1 β N N N [x] N a vector of length N with all entries x

36 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 36 We can rearrange the PageRank equation r = ββ r + 1 β N N [(1-β)/N] N is an N-vector with all entries (1-β)/N M is a sparse matrix! 10 links per node, approx 10N entries So in each iteration, we need to: Compute r new = β M r old Add a constant value (1-β)/N to each entry in r new

37 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 37 Encode sparse matrix using only nonzero entries Space proportional roughly to number of links Say 10N, or 4*10*1 billion = 40GB Still won t fit in memory, but will fit on disk source node degree destination nodes 0 3 1, 5, , 64, 113, 117, , 23

38 Assume enough RAM to fit r new into memory Store r old and matrix M on disk Then 1 step of power-iteration is: Initialize all entries of r new to (1-β)/N For each page p (of out-degree n): Read into memory: p, n, dest 1,,dest n, r old (p) for j = 1 n: r new (dest j ) += β r old (p) / n r new src degree destination 0 3 1, 5, , 64, 113, , 23 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 38 r old

39 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 39 Assume enough RAM to fit r new into memory Store r old and matrix M on disk In each iteration, we have to: Read r old and M Write r new back to disk IO cost = 2 r + M Question: What if we could not even fit r new in memory?

40 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets r new src degree destination 0 4 0, 1, 3, , , 4 r old

41 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 41 Similar to nested-loop join in databases Break r new into k blocks that fit in memory Scan M and r old once for each block k scans of M and r old k( M + r ) + r = k M + (k+1) r Can we do better? Hint: M is much bigger than r (approx 10-20x), so we must avoid reading it k times per iteration

42 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets r new src degree destination 0 4 0, r old

43 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 43 Break M into stripes Each stripe contains only destination nodes in the corresponding block of r new Some additional overhead per stripe But it is usually worth it Cost per iteration M (1+ε) + (k+1) r

44 2/7/2012 Jure Leskovec, Stanford C246: Mining Massive Datasets 44 Measures generic popularity of a page Biased against topic-specific authorities Solution: Topic-Specific PageRank (next) Uses a single measure of importance Other models e.g., hubs-and-authorities Solution: Hubs-and-Authorities (next) Susceptible to Link spam Artificial link topographies created in order to boost page rank Solution: TrustRank (next)