key key total j d keys closer to root float wt ivo vn
play

. ( key key - total j 'D keys closer to root float wt Ivo , . . .vn - PowerPoint PPT Presentation

- , Implementation :( as extended BST ) weight Balance Overview : : Internal node : stores - splay trees - Given asetof keys - Static X={ Xo , Optimality . ,xn . , } More frequently accessed . . splitter - and values . ( key key -


  1. - , Implementation :( as extended BST ) weight Balance Overview : : Internal node : stores - splay trees - Given asetof keys - Static X={ Xo , Optimality . ,xn . , } More frequently accessed . . → splitter - and values . ( key key - → total j 'D keys closer to root float wt Ivo , . . .vn . , ) ⇒ weight - and weights . - balanced weight of ax xx Tv - two , entries in subtree . ,wn . , } , right -1 trees 1 $ . . . left - Assume : iesttiesataedwm . . . External Node : sorted syoaii.ea.awei.I.iq - sxn . , xosx ,s . . . ' key key , ④ positivity Wi > o . - probability pseudo : wi total weight ¥ 5 - Let : = Given Yeast ,aBsT - Let :pi=WiFw How to ( Nearly ) Achieve a - balanced if pseudo - prob is - Obs :& ? ; ! ! } } Shannon 's bound ⇒ discrete for all internal nodes p . - balanced . . . → Weight prob balance ( p ) sa distribution . tree : Perfectly balanced 9=42 Shannon 's Theorem : Ifpi { tathekastfropyipoiflgj ! P .in#alledT/balancelpg=maxlwtlp.lettl.wt-p.rightDa--4s :b reasonable → For each node p is : : Arbitrarily bad wtlp ) - total weight . of accessing Xi the prob =L 's subtree , of keys nip any Bst has expected search compromise wtlp ) .-wt .

  2. - - ideal Ax Balance by Rebuilding ksb How to maintain balance ? Example :oi_ : key :%/I µµ /-% : utile an array Alo . Option 's . let ] Given : - Rotations : Similar to of external nodes : .is/ltnt4H3t2t4YJ.wsY6Ali3.wt . key AVL trees ( singles double ) Ali ] , Ali ) - value . Dm b - → BBK ) trees =3 . . - Rebuild subtrees : Similar t.buifralalo.zDR-buildtreelAG.is 'D - Assume Keys are sorted to scapegoat - Assume weights get > o return I . . $ 9 ⑤ . m . . - saetestrgetiteerigtodm.in#emi:neahTreesEmEt . ' weight Q ① - Balanced - based median : weight add a :O - K - D) i build Tree ( Alo ← Total wt - Let Ali ] .wt .az#:nne*/-Goal:Splitatosjskthal- = . iflk-tretvrna.co/x.basecase*loT--EiijAli3.wt ? jt ← o ¥ %÷tj , ] ' - Iii : - Letwin Iii A total weight 't - . - O ; Rwtsto ;Dm , IT " is "÷÷÷÷÷÷÷ Init : b=o ; Lwt - . . . . . . . . %fbfifgffemgyn.to ? j.iijjAmmi=D3 :* ¥ minimizes : ← Most balanced split - org Aj Amin - . t-T est - buildtreelacb.k.is ) !? CEi " R Q - A new Int Node ( Alb ] . key ,L,R ) #Esphthe return .it .

  3. - , ⇐ Analysis Bad weight distributions ? Butitis pretty close ! : Does this algorithm produce - Ifa weight is very large the optimal tree Curt . expected relative to neighbors , Theorem :( Mehlhorn 'M ) rebalance maybe ineffective The above balanced split case search time ) ? algorithm produces a tree - ÷÷ ⇒ µ - The optimal BST can be " " d - No . search time is → whose exp . computed by dynamic programming t ⑧ SH -13 are ) weights Lemma : - entropy bound " ( not too much :c :* :* : nisig.EE ? a-n7w7 - . . says , where " H Lionizing , nice the . variation ) , inserts delete .tn :c - Jackhammer rebuilding . - → Checks Rebuild : . - * Ifnosuohp found Trees - when returning - ÷ ÷ from Great ! Tree is balanced Find : Same usual . Tree as : heights log%n,soO4ogn ) . : : Jackhammer ! recursive calls , update Else : ÷ time - guaranteed each node 's weight . . . 's subtree i - Traverse p ; . p.wt-p.left.wttp.right.int in order , store extern : : Insert Delete : Start same - starting at root , walk down nodes in array Alak - D i : as - Replace p 's subtree : search path standard BST . Stop at first i ! → After operation completes node pst ' with qq.gr ? er..-.s.ba1anoecp , > g÷9÷÷} ! I ¥ 7÷i i . checks rebuild buildtreeln ) " " " ° ,

  4. - , when to rebuild ? Dictionary Operations : f÷÷÷÷÷÷÷÷i very heavy entries : 's weight an entry - If - find - when " backing out " is too high as usual , rebuilding : from insertldelete - insert : insert as usual is ineffective update node weights ynutbarlgniidnig , - Example but rebuild if needed - walk down search - delete doesn't help ! : delete as usual , path from root but rebuild if needed " :÷÷÷÷÷÷÷÷÷÷÷i÷ . ÷ - and - - balance : Every ( a. P ) : maxcpt For node p - ratio ( p )sp max max weightings internal node p is either 's subtree then : . - balanced . maxlp ) x on - exempt - Rebuild p - ratio ( pls p Max - : weight ( P ) - Traverse pisubtreeinorder For any set of Lemma : ' Given parameter Opal - collect external nodes in weighted entries , Jan , . anmodae.irsafioefsm.pt 't .fi ? idtreeunp--Y4 - remain : 'm ° ,

Download Presentation
Download Policy: The content available on the website is offered to you 'AS IS' for your personal information and use only. It cannot be commercialized, licensed, or distributed on other websites without prior consent from the author. To download a presentation, simply click this link. If you encounter any difficulties during the download process, it's possible that the publisher has removed the file from their server.

Recommend


More recommend