Presentation is loading. Please wait.

Presentation is loading. Please wait.

Kinetic data structures. Goal Maintain a configuration of moving objects Each object has a posted flight plan (this is essentially a well behaved function.

Similar presentations


Presentation on theme: "Kinetic data structures. Goal Maintain a configuration of moving objects Each object has a posted flight plan (this is essentially a well behaved function."— Presentation transcript:

1 Kinetic data structures

2 Goal Maintain a configuration of moving objects Each object has a posted flight plan (this is essentially a well behaved function of time)

3 Example 1 Maintain the closest pair among points moving in the plane

4 Example 2 Maintain the convex hull of points moving in the plane

5 Elements of a KDS An event queue (A heap of discrete times) The event queue will contains all times where the combinatorial structure of the configuration may change Like a “sweep” of the time dimension

6 Example 3 Maintain the topmost among points moving along the y-axis

7 Look at the ty-plane t y

8 We are interested in the upper envelope t y

9 Solution Calculate this upper envelope ! Sharir, Hart, Agarwal and others: –The complexity of the envelope is close to linear if any pair of function intersect at most s times –Can compute it in O(n log(n)) time

10 Problem If we would like to change a trajectory then we need to recompute te envelope That takes O(nlog(n)) time We want to be able to change a trajectory faster

11 Another solution Maintain the points sorted For every pair of points put in the event queue the time when they switch order

12 Example 3

13 Problem We process Ω(n 2 ) events But the configuration changes only linear (or close to linear) number of times…

14 So what do we want from a KDS to be good You maintain a set of certificates that as long as they are valid the configuration does not change. Want: The number of times a certificate fails (internal events) to be small relative to the number of times the configuration changes (external events)  Efficient

15 So what do we want from a KDS to be good (Cont) Process a certificate failure fast  responsive Small space  compact Object participates in a small # of cetificates (can change trajectories easily)  local

16 Dynamic KDS Want also to be able to insert and delete objects efficiently

17 So what would be a good solution for this problem ? Maintain the topmost among points moving along the y-axis

18 A tournament tree a b c d a b c d

19 a b c d a b c d d c d

20 a b c d a b c d d c d For each internal node maintain in an event queue the next time where the children flip order

21 a b c d a b c d d c d Processing of an event: Replace the winner and replace O(log(n)) events in the event queue t y Takes O(log 2 (n)) time  responsive Linear space  compact Each point participates in O(log n) events  local

22 a b c d   r a b c d d c d What is the total # of events ? t y Events at r correpond to changes at the upper envelope, lets say there are O(n) Events at 1 correponds to change at the upper envelope of {b d}  O(n/2) … In total we get O(nlog(n)) events  efficient

23 a b c d   r a b c d d c d Handeling insertions/deletions ? t y Use some kind of a balanced binary search tree Each node charges its events to the upper envelope of its subtree Without rotations we get O(nlog(n)) events

24 a b c d   r a b c d d c d Handeling insertions/deletions t y Because of rotations each point participates in more than O(log n) envelopes Use a BB[alpha] tree  think of each pair of nodes participating in a rotation as new nodes, then the total size of envelopes corresponding to new nodes is O(nlog(n))

25 a b c d   r a b c d d c d t y We’ll focus now a bit more at the case where the points move with constant velocity Can redefine the problem so we do not insist on maintaining the upper envelope explicitly at all times

26 A collection of items, each with an associated key. key (i) = a i x + b i a i,, b i reals, x a real-valued parameter a i = slope, b i = constant Operations: make an empty heap. insert item i with key a i x + b i into the heap: insert(i,a i,b i ) find an item i of minimum key for x = x 0 : find-max( x 0 ) delete item i : delete(i) Parametic Heap

27 A parametric heap such that successive x-values of find maxs are non-decreasing. (Think of x as time.) x c = largest x in a find max so far (current time) Additional operation: increase the key of an item i, replacing it by a key that is no larger for all x  x c : increase-key(i,a,b) Kinetic Heap

28 Equivalent problems: maintain the upper envelope of a collection of lines in 2D projective duality maintain the convex hull of a set of points in 2D under insertion and deletion What is known about parametric and kinetic heaps?

29 Overmars and Van Leeuwen (1981) O( log n) time per query O(log 2 n) time per update, worst-case Chazelle (1985), Hershberger and Suri (1992) (deletions only) O( log n) time per query, worst-case O(n log n) for n deletions Results I

30 Results II Chan (1999) Dynamic hulls and envelopes O( log n) time per query O(log 1+  n) time per update, amortized Brodal and Jacob (2000), Kaplan, Tarjan, Tsioutsiouliklis (2000) O( log n) time per query O( log n log log log n) time per insert, O( log n log log n) timer per delete, amortized

31 Results III Basch, Guibas, and Hershberger (1997) “Kinetic” data structure paradigm

32 Users One server, many possible items to send (say, all the same length) One broadcast channel. Users submit requests for items. Goal: Satisfy users as well as possible, making decisions on-line. (say, minimize sum of waiting times) Server: many data items Broadcast channel (single-item) Broadcast Scheduling

33 Greedy = Longest Wait first (LWF): Send item with largest sum of waiting times. R x W : send item with largest ( # requests x longest waiting time) Scheduling policies (vs. number of requests or longest single waiting time)  

34 34 Results of Mike Franklin and others: LWF schedules well “in practice” (in simulations) but too expensive (linear-time) This claim used to justify approximations to R x W, still linear-time but with a smaller (parameterized) constant.

35 Questions (for an algorithm guy or gal) LWF does well compared to what? Try a competitive analysis Can we improve the cost of LWF?   What data structure? Open question 1 Will talk about this

36 Need a max-heap (replace find min by find max, decrease key by increase key, etc) Can implement LWF or R x W or any similar policy: Broadcast decision is find max plus delete Request is insert (if first) or increase key (if not) Only find max need be real-time, other ops can proceed concurrently with broadcasting Slopes are integers that count requests Broadcast scheduling via kinetic heap

37 LWF: Suppose a request for item i arrives at time t s If i is inactive then insert(i, t-t s ) If i is active with key at+b then increase-key(i, (a+1)t+(b-t s )) To broadcast an item at time t s we perform delete-max(t s ) and broadcast the item returned. Broadcast scheduling via kinetic heap (Cont.)


Download ppt "Kinetic data structures. Goal Maintain a configuration of moving objects Each object has a posted flight plan (this is essentially a well behaved function."

Similar presentations


Ads by Google