Optimal Sampling for Sliding Windows课件.ppt

Optimal Sampling for Sliding Windows;Data Streams: Formal Definition;p1;p1;p1;p1;Random Sampling;Problem Definition;Our Results: Summary;;Motivation;Why Random Sampling?;Sampling methods;Why sliding windows?;Some papers on SW ;Discussion and Related Work Why na?ve methods fail?Why previous methods are not deterministic?  ;Why the problem is not trivial?;p1;Why Periodic Sampling cannot be used?;Oversampling;Babcock, Datar, Motwani, SODA 2002.;Our Main Idea (informal);Sampling With Replacement for Sequence-Based Windows ;Expired element;How to maintain a single sample;pN-5;Why it works?;pN-5;pN-5;Sampling Without Replacement for Sequence-Based Windows ;The Algorithm ;pN-5;Sampling Without Replacement for Timestamp-Based Windows ;Black-Box Reduction ;Black Box reduction:the key idea;Cascading Lemma (Illustration);What we did not cover (see the paper):Sampling With Replacement for Timestamp-Based Windows ;;Implications of our Results;Conclusions;BACKUP SLIDES;Sampling With Replacement for Timestamp-Based Windows ;Timestamp-based window;A;A;A;A;The First Problem;Solving Problem 1: Covering Decomposition;Covering Decomposition : implementation ;A;Solving Problem 2: Generating Implicit Events;Why we can generate implicit events?;A;Proof of Lemma 1: Constructing Y;Proof of Lemma 1:Correctness;Proof of Lemma 1:Constructing Y;Lemma 2: telescope;Conclusions;Thank you!