Theoretical Computer Science. Approximately optimal trees for group key management with batch updates

Size: px
Start display at page:

Download "Theoretical Computer Science. Approximately optimal trees for group key management with batch updates"

Transcription

1 Theoreical Compuer Science ) Conens liss available a ScienceDirec Theoreical Compuer Science journal homepage: Approximaely opimal rees for group key managemen wih bach updaes Minming Li a,, Ze Feng a, Nan Zang b, Ronald L. Graham b, Frances F. Yao a a Deparmen of Compuer Science, Ciy Universiy of Hong Kong, Hong Kong b Deparmen of Compuer Science and Engineering, Universiy of California a San Diego, Unied Saes a r i c l e i n f o a b s r a c Keywords: Key managemen Key rees Bach updae Approximaion algorihms Opimaliy We invesigae he group key managemen problem for broadcasing applicaions. Previous work showed ha, in handling key updaes, bach rekeying can be more cos effecive han individual rekeying. One model for bach rekeying is o assume ha every user has probabiliy p of being replaced by a new user during a bach period wih he oal number of users unchanged. Under his model, i was recenly shown ha an opimal key ree can be consruced in linear ime when p is a consan and in On 4 ) ime when p 0. In his paper, we invesigae more efficien algorihms for he case p 0, i.e., when membership changes are sparse. We design an On) heurisic algorihm for he sparse case and show ha i produces a nearly 2-approximaion o he opimal key ree. Simulaion resuls show ha is performance is even beer in pracice. We also design a refined heurisic algorihm and show ha i achieves an approximaion raio of 1 + ɛ for any fixed ɛ > 0 and n, as p 0. Finally, we give anoher approximaion algorihm for any p 0, 0.69) which is shown o be quie good by our simulaions Elsevier B.V. All righs reserved. 1. Inroducion Wih he increase of subscripion-based nework services, sraegies for achieving secure mulicas in neworks are becoming more imporan. For example, o limi he service access o he auhorized subscribers only, mechanisms such as conen encrypion and selecive disribuion of decrypion keys have been found useful. One can regard he secure mulicas problem as a group broadcas problem, where we have n subscribers and a group conroller GC) ha periodically broadcass messages e.g., a video clip) o all subscribers over an insecure channel. To guaranee ha only he auhorized users can decode he conens of he messages, he GC will dynamically mainain a key srucure for he whole group. Whenever a user leaves or joins, he GC will generae some new keys as necessary and noify he remaining users of he group in some secure way. Surveys of key managemen for secure group communicaions can be found in [1,2]. In his paper, we consider he key ree model [] for he key managemen problem. We describe his model briefly as follows a precise formulaion is given in Secion 2). Every leaf node of he key ree represens a user and sores his individual key. Every inernal node sores a key shared by all leaf descendans of ha inernal node. Every user possesses all he keys along he pah from he leaf node represening he user) o he roo. To preven revoked users from knowing fuure message This work was suppored in par by he Naional Basic Research Program of China Gran 2007CB807900, 2007CB807901, he US Naional Science Foundaion Gran CCR , grans from he Research Grans Council of Hong Kong under Projec Number CiyU , 1165/04E and a gran from Ciy Universiy of Hong Kong Projec no ). Corresponding auhor. addresses: minmli@cs.ciyu.edu.hk M. Li), fengze@cs.ciyu.edu.hk Z. Feng), nzang@cs.ucsd.edu N. Zang), graham@ucsd.edu R.L. Graham), csfyao@ciyu.edu.hk F.F. Yao) /$ see fron maer 2008 Elsevier B.V. All righs reserved. doi: /j.cs

2 1014 M. Li e al. / Theoreical Compuer Science ) conens and also o preven new users from knowing pas message conens, he GC updaes a subse of keys, whenever a new user joins or a curren user leaves, as follows. As long as here is a user change among he leaf descendans of an inernal node v, he GC will 1) replace he old key sored a v wih a new key, and 2) broadcas o all users) he new key encryped wih he key sored a each child node of v. Noe ha only users corresponding o he leaf descendans of v can decipher useful informaion from he broadcas. Furhermore, his procedure mus be done in a boom-up fashion i.e., saring wih he lowes v whose key mus be updaed see Secion 2 for deails) o guaranee ha a revoked user will no know he new keys. The cos of he above procedure couns he number of encrypions used in sep 2) above or equivalenly, he number of broadcass made by he GC). When users change frequenly, he mehod for updaing he group keys whenever a user leaves or joins may be oo cosly. Thus, a bach rekeying sraegy was proposed by Li e al. in [4], whereby rekeying is done only periodically insead of immediaely afer each membership change. I was shown by simulaion ha among he oally balanced key rees where all inernal nodes of he ree have branching degree 2 i ), degree 4 is he bes when he number of requess leave/join) wihin a bach is no large. For a large number of requess, a sar a ree of deph 1) ouperforms all such balanced key rees. Furher work on he bach rekeying model was done by Zhu e al. in [5]. They inroduced a new model where he number of joins is assumed o be equal o he number of leaves during a bach updaing period and every user has probabiliy p of being replaced by a new user for some p. They sudied he opimal ree srucure subjec o wo resricions: A) he ree is oally balanced, and B) every node on level i has 2 k i children for some parameer k i depending on i. Under hese resricions, characerizaions of he opimal key ree were given ogeher wih a consrucion algorihm. Recenly, Graham, Li, and Yao [6] sudied he srucure of he rue opimal key ree when resricions A) and B) are boh removed. They proved ha, when p > 1 1/ 0.07, he opimal ree is an n-sar. When p 1 1/, hey proved a consan upper bound 4 for he branching degree of any inernal node v oher han he roo, and also an upper bound of 4/ log q, where q = 1 p, for he size of he subree rooed a v. By using hese characerizaions, hey designed an On 4 ) algorihm for compuing he opimal key ree for n users. The running ime of heir algorihm is in fac linear when p is a fixed consan, bu becomes On 4 ) when p approaches 0. Alhough polynomial, he On 4 ) complexiy is sill oo cosly for large scale applicaions. Indeed, he case p 0 when user changes are sparse) is a realisic scenario in many applicaions. In his paper, we invesigae more efficien heurisics for he sparse case. As shown in [6], degree is quie favored in he opimal ree as p 0. In fac, heir resuls implied ha, for n =, he opimal key ree is a balanced ernary ree, and for many oher values of n, he opimal ree is as close o a balanced ernary ree wih n leaves as possible, subjec o some number-heoreical properies of n. In his paper, we invesigae how closely a simple ernary ree can approximae he opimal ree for arbirary n. We propose a heurisic LR which consrucs a ernary ree in a lef o righ manner and prove ha i gives a nearly 2- approximaion o he opimal ree. Simulaion resuls show ha he heurisic performs much beer han he heoreical bound we obain. We hen design a refined heurisic LB whose approximaion raio is shown o be 1 + ɛ as p 0 where ɛ can be made arbirarily small. We also generalize he heurisic LR o a heurisic GLR for he case when p 0, 1 1/ ). The res of he paper is organized as follows. In Secion 2, we describe he bach updae model in deail. In Secion, we esablish a lower bound for he opimal ree cos as p 0. The lower bound is useful for obaining performance raios for our approximaion algorihms. In Secion 4, we describe he heurisic LR and analyze is performance agains he opimal key ree; some simulaion resuls are also given. Then we design a refined heurisic LB and analyze is performance in Secion 5. In Secion 6, we generalize he heurisic LR for p in a more general range, and show ha he generalized LR GLR) performs well by simulaion resuls. Finally, we summarize our resuls and menion some open problems in Secion Preliminaries Before giving a precise formulaion of he key ree opimizaion problem o be considered, we briefly discuss is moivaion and review he basic key ree model for group key managemen. This model is referred o in he lieraure eiher as a key ree [] or LKH logical key hierarchy) [7]. In he key ree model, here are a Group Conroller GC), represened by he roo, and n subscribers or users) represened by he n leaves of he ree. The ree srucure is used by he GC for key managemen purposes. Associaed wih every node of he ree wheher inernal node or leaf) is an encrypion key. The key associaed wih he roo is called he Traffic Encrypion Key TEK), which is used by he subscribers for accessing encryped service conens. The key k v associaed wih each nonroo node v is called a Key Encrypion Key KEK), which is used for updaing he TEK when necessary. Each subscriber possesses all he keys along he pah from he leaf represening he subscriber o he roo. In he bach updae model o be considered, only simulaneous join/leave is allowed; ha is, whenever here is a revoked user, a new user will be assigned o ha vacan posiion. This assumpion is jusified since, in a seady sae, he number of joins and deparures would be roughly equal during a bach processing period. To guaranee forward and backward securiy, a new user assigned o a leaf posiion will be given a new key by he GC and, furhermore, all he keys associaed wih he ancesors of he leaf mus be updaed by he GC. The updaes are performed from he lowes ancesor upward for securiy reasons. We hen explain he updaing procedure ogeher wih he updaing cos in he following. The GC firs communicaes wih each new subscriber separaely o assign a new key o he corresponding leaf. Afer ha, he GC will broadcas cerain encryped messages o all subscribers in such a way ha each valid subscriber will know all he new keys associaed wih is leaf-o-roo pah while he revoked subscribers will no know any of he new keys. The

3 M. Li e al. / Theoreical Compuer Science ) GC accomplishes his ask by broadcasing he new keys, in encryped form, from he lowes level upward as follows. Le v be an inernal node a he lowes level whose key needs o be bu has no ye been) updaed. For each child u of v, he GC broadcass a message conaining E k newk new u v ), which means he encrypion of k new v wih he key k new u. Thus he GC sends ou d v broadcas messages for updaing k v if v has d v children. Updaing his way ensures ha he revoked subscribers will no know any informaion abou he new keys as long as hey do no know he new key k new u in a lower level, hey canno ge he informaion of he new key k new v in a higher level) while curren subscribers can use one of heir KEKs o decryp he useful E k newk new u v ) sequenially unil hey ge he new TEK. We adop he probabilisic model inroduced in [5] ha each of he n posiions has he same probabiliy p o independenly experience subscriber change during a bach rekeying period. Under his model, an inernal node v wih N v leaf descendans will have probabiliy 1 q N v ha is associaed key k v requires updaing, where q = 1 p. The updaing incurs d v 1 q N v ) expeced broadcas messages by he procedure described above. We hus define he expeced updaing cos CT) of a key ree T by CT) = v d v 1 q N v ), where he sum is aken over all he inernal nodes v of T. I is more convenien o remove he facor d v from he formula by associaing he weigh 1 q N v wih each of v s children. This way we express CT) as a node weigh summaion: for each non-roo ree node u, is node weigh is defined o be 1 q N v, where v is u s paren. The opimizaion problem we are ineresed in can now be formulaed as follows. Opimal key ree for bach updaes: We are given wo parameers, 0 p 1 and n > 0. Le q = 1 p. For a rooed ree T wih n leaves and node se V including inernal nodes and leaves), define a weigh funcion wu) on V as follows. Le wr) = 0 for roo r. For every non-roo node u, le wu) = 1 q N v, where v is u s paren. Define he cos of T as CT) = u V wu). Find a T for which CT) is minimized. We say ha such a ree is p, n)-opimal, and denoe is cos by OPTp, n).. Lower bound for opimal ree cos as p 0 In a ree T wih n leaves, denoe he se of leaf nodes as LT), and for each leaf u, le he se of ancesor nodes of u including u iself) be denoed by Ancu). To obain a lower bound for he opimal ree cos, we firs rewrie CT) as CT) = u V wu) = u V N u wu) N u = u LT) x Ancu) wx) = cu), N x where we define cu) = wx) x Ancu). In oher words, we disribue he weigh N x wu) associaed wih every node u V evenly among is leaf descendans, and hen sum he cos over all he leaves of T. Le he pah from a leaf u o he roo r be p 0 p 1... p k 1 p k, where p 0 = u and p k = r. Noe ha cu) = k 1 i=0 u LT) 1 q Np i+1 N pi uniquely deermined by he sequence of numbers {N p0, N p1,..., N pk }, where N p0 = 1 and N pk = n. We will hus exend he definiion of c o all such sequences {a 0, a 1,..., a k }, and analyze he minimum value of c. Definiion 1. Le S n denoe any sequence of inegers {a 0, a 1,..., a k } saisfying 1 = a 1 < a 2 < < a k = n. We call S n an n-progression. Define cp, S n ) o be cp, S n ) = k 1 q a i i=1, and le a i 1 Fp, n) be he minimum of cp, S n ) over all n-progressions S n. For n =, he special n-progression {1,, 9,..., 1, } will be denoed by S n. is Thus, we have CT) n Fp, n) for any ree T wih n leaves, and hence OPTp, n) n Fp, n). Nex we focus on properies of cp, S n ) and Fp, n). Firs, we derive he following monoone propery for Fp, n). 1) Lemma 2. Fp, n) < Fp, n + 1). Proof. Suppose S n+1 = {1, a 1, a 2,..., a k 1, n + 1} is he opimal n + 1)-progression ha achieves he value Fp, n + 1). Le S n = {1, a 1, a 2,..., a k 1, n}. Because 1 qn+1 a k 1 > 1 qn, we know ha a k 1 cp, S n+1 ) > cp, S n ). By definiion, we have Fp, n) cp, S n ). Combining hese wo facs, we have Fp, n) < Fp, n + 1). For a given n-progression S n = {1, a 1, a 2,..., a k 1, n}, he slope of cp, S n ) a p = 0 is denoed by λ Sn and can be expressed as λ Sn = k 1 a i+1 i=0, where a a 0 i = 1 and a k = n. The minimum cp, S n ) as p 0 will be achieved by hose S n wih cp, S n ) having he smalles slope a p = 0. We nex prove he following lemma. Lemma. When n =, he n-progression S n = {1,, 9,..., 1, } saisfies he following relaion: λ Sn λ S n for any S n. Proof. For posiive numbers b 1, b 2,..., k b k, we have i=1 b i k k i=1 b i) 1 k. Therefore, λsn kn 1 k if Sn consiss of k numbers. We now esimae a lower bound for f k) = kn 1 k when n =. Consider gk) = log f k) = ln k +. Noice ln k ha g k) = 1. Therefore, we have k ln k 2 g k) < 0 when k < ln and g k) > 0 when k > ln. This implies

4 1016 M. Li e al. / Theoreical Compuer Science ) Fig. 1. Trees generaed by LR for n = 2 9. ha gk), and hence f k), is minimized when k = ln. Therefore, we have f k) f ln ) = 1 ln ln, which implies λ Sn 1 ln ln. On he oher hand, we know ha λs n = f ) =. Hence we have λ Sn f ln ) 1+ln ln = ln λ S n f ) We obain he following heorem from he above analysis. Theorem 4. For n < +1, we have OPTp, n) n cp, S ) when p 0. Proof. This is a direc consequence of Lemmas 2 and and inequaliy 1). 4. Heurisic LR and is approximaion raio We design he heurisic LR as follows. LR mainains an almos balanced ernary ree i.e., he deph of any wo leaves differ by a mos 1) in which a mos one inernal node has degree less han. Moreover, LR adds new leaves incremenally in a lef o righ order. Fig. 1 shows he ree we ge by using LR for n = 2,..., 9. We can also recursively build a key ree using LR in he following way. For a ree wih n leaves, he number of leaves in he roo s hree subrees is decided by he able below, while for a ree wih 2 leaves, he ree srucure is a roo wih wo children a sar). No. of leaves Lef) Middle) Righ) n < 5 1 n n < 7 1 n n < +1 n 2 We denoe he ree wih n leaves consruced by LR as T n. Noe ha his heurisic only needs linear ime o consruc a ernary ree. Furhermore, he srucure of he ernary ree can be decided in log n ime, because, every ime we go down he ree, here is a mos one subree whose number of leaves is no a power of and needs furher calculaion. Le LRp, n) denoe he cos of he ernary ree consruced by LR for given n and p. To obain an upper bound for LRp, n), we firs prove he following lemmas. Lemma 5. The inequaliy LRp, n) < LRp, n + 1) holds for all n > 0 and 0 < p < 1. Proof. We view CT n ) as he node weigh summaion given in Secion 2 and compare he cos of he corresponding nodes wu) and wu ) in T n and T n+1, respecively. Due o he addiion of one leaf node, if wu) = 1 q k, hen wu ) = wu) or wu ) = 1 q k+1. Therefore we have wu) wu ). There are also addiional weighs associaed wih nodes ha appear in T n+1 bu no in T n. This proves he lemma. Lemma 6. For any ineger > 0 and 0 < q < 1, we have 1 q > 1 q Proof. Noe ha i=1 qi 1 > 1 + q + q 2 ) i=1 qi 1 = +1 i=1 qi 1. The lemma is proved by muliplying 1 q on boh sides. Lemma 7. For any ineger > 0 and 0 < q < 1, we have LRp, +1 ) < ) LRp, ). Proof. By Lemma 6 and he definiions, we have cp, S )/ > cp, S )/ + 1). Therefore, we have +1 LRp, +1 ) = +1 cp, S ) < ) cp, S ) = ) LRp, ). Now we are ready o prove he firs approximaion raio. Theorem 8. When p 0, we have LRp, n) < )OPTp, n).

5 M. Li e al. / Theoreical Compuer Science ) Proof. Suppose n < +1. We claim he following: LRp, n) < LRp, +1 < ) LRp, ) = ) cp, S ) ) OPTp, n). The firs inequaliy is implied by Lemma 5 and he second one by Lemma 7. The las inequaliy holds due o Theorem 4. In he above discussion, we use he smalles balanced ernary ree wih no fewer han n leaves as an upper bound for LRp, n). By adding a small number of leaves insead of filling he whole level, we can obain a beer approximaion raio, which is shown below. We divide he inegers in he range, +1 ] ino hree consecuive subses of equal size H = +1 as follows: P 1 =, + H], P 2 = + H, + 2H], P = + 2H, +1 ]. For any n P i, we can use LRp, n ), where n = max P i o upper bound he value of LRp, n) by Lemma 5. Le = LRp, ) LRp, 1 ) and define a = 1 q +1. Noice ha LRp, +1 ) = a + LRp, ) = a + + LRp, 1 ). I is no hard o verify he following inequaliies based on he definiion of he ree cos: LRp, 7 1 ) < LRp, +1 ), LRp, 5 1 ) < LRp, +1 ) 2. We now derive a lower bound for he value of. Lemma 9. For 0 < p < 1, we have 1 6 LRp, +1 ). Proof. We only need o prove > a + LRp, 1 ). By he definiion of, we know ha = 2 LRp, 1 ) + 1 q ). Then by using Lemma 6, we have 1 q ) 1 q +1 ), which implies LRp, 1 ) + 1 q ) 1 q +1 ). Therefore, we have = 2 LRp, 1 ) + 1 q ) > LRp, 1 ) + a. By making use of Lemma 9, we can obain he following heorem on he performance of LR. Theorem 10. When p 0, we have LRp, n) < )OPTp, n). Proof. We prove he heorem using Lemma 9 and similar argumens used in Theorem 8. The discussion below is divided ino hree cases according o he value of n. Case A). < n 5 1. LRp, n) < LRp, 5 1 ) < 2 LRp, +1 ) < ) OPTp, n) n ) OPTp, n). Case B). 5 1 < n 7 1. LRp, n) < LRp, 7 1 ) < 5 6 LRp, +1 ) < ) OPTp, n) n < ) OPTp, n) 5 ) 1 < OPTp, n).

6 1018 M. Li e al. / Theoreical Compuer Science ) a) p = 0.1 and p = b) p = 0.01/n. Fig. 2. Simulaion resuls on Raiop, n) for p = 0.1, p = 0.01 and p = 0.01/n. Fig.. Simulaion resuls for fixed n. Case C). 7 1 < n +1. LRp, n) < LRp, +1 ) < ) OPTp, n) n < ) OPTp, n) 7 ) 1 < OPTp, n). We run simulaions on he performance of LR for various values of p and n. Define Raiop, n) = LRp, n)/optp, n). Fig. 2a) shows Raiop, n) as a funcion of n for p = 0.1 and p = 0.01 respecively. Wihin he simulaion range, we see ha, for he same n, Raiop, n) is larger when p is larger. For p = 0.1 he maximum raio wihin he range is below We hen simulaed he performance of LR when p 0 as a funcion of n. Fig. 2b) shows Raiop, n) when we se p = 0.01/n. We found he maximum raio reached wihin he simulaion range o be less han Fig. plos Raiop, n) as a funcion of p while n is chosen o be n = 100 and n = 500, respecively. Noice ha each curve has some dips and is no monoonically increasing wih p. 5. Heurisic LB and is approximaion raio In his secion, we consider a more refined heurisic LB Level Balance) and show ha i has an approximaion raio of 1 + ɛ as p 0 for fixed n). The heurisic LB adops wo differen sraegies for adding a new leaf o a ree when n grows from o +1. Le T denoe he ernary ree wih n leaves consruced by LB. When < n 2, he heurisic LB changes he lefmos leaf on level in T 1 ino a 2-sar i.e., an inernal node wih wo leaf children); when 2 < n +1, he heurisic LB changes he lefmos 2-sar on level in T 1 ino a -sar an inernal node wih hree leaf children). In he following discussion, we refer o he iniial formulaion of CT) as a node weigh summaion over all nodes: CT) = u V wu). Define he slope of CT) a p = 0 as and le =. When T changes incremenally o

7 M. Li e al. / Theoreical Compuer Science ) Fig. 4. Approximae ree srucure generaed by GLR when n = 0 p = 0.1. T, where < n 2, every ime he size is increased by 1, we change a single leaf node o a 2-sar. Therefore, in all inermediae levels excep for he roo and he boom level), exac hree nodes will undergo a weigh change: i changes from 1 q N v o 1 q Nv+1. For he wo newly added nodes, he weigh of each node is 1 q 2. Alogeher, hese changes conribue an increase of o he slope. When T 2 is changed incremenally o T, where 2 < n +1, by using a similar argumen, we conclude ha every ime he size is increased by 1, all weigh changes ogeher conribue an increase of = + 5 o he slope, where 9 4 means he slope increase due o 1 q ) 21 q 2 ) because we change a 2-sar o a -sar). To summarize, our new heurisic has he following propery. = { βt n if < n 2, n if 2 < n +1. Using he base value of and he recurrence relaion 2), we have he following lemma. Lemma 11. { + 4)n 4 if = < n 2, + 5)n 6 if 2 < n +1. 2) By Lemma 2 and Theorem 4, we can prove he following heorem. Theorem 12. When p 0, we have CT ) < n )OPTp, n). Proof. To compare he cos of wo rees as p 0 is in fac comparing slopes of he corresponding CT) a p = 0. Noe ha, for a full ernary ree T wih heigh, he slope of CT) a p = 0 equals. Always le = in he following. Using Lemma 4, we can prove he heorem by proving < ) n as follows. Case A). n = + r, where 0 < r. In his case, we have = + r + 4) < + r + ) + = + ) + r) = ) n. Case B). n = 2 + r, where 0 < r. In his case, we have = + + 4) + r + 5) = r + 5r r + r = + )2 + r) = ) n. The upper bound in Theorem 12 can be furher improved o 1+ɛ, where ɛ can be made arbirarily small when p 0. To accomplish his, we make use of he following echnical lemma, which was originally proved in [6] for a differen purpose. Lemma 1. For any ree T wih n leaves, we have { + 4)n 4 if < n 2, + 5)n 6 if 2 < n +1. By comparing Lemmas 11 and 1 we can deduce ha, for he ree T consruced by he heurisic LB, he slope of CT ) is equal o he lower bound of he slope of he opimal key ree wih n leaves. Therefore, as p 0, he approximaion raio of LB can be arbirarily close o 1. This leads o he following heorem. Theorem 14. For any fixed n, we have CT ) < 1 + ɛ)optp, n) when p 0.

8 1020 M. Li e al. / Theoreical Compuer Science ) a) p = 0.. b) p = 0.2. c) p = 0.1. d) p = 0.01, he raio of LR and GLR. Fig. 5. Simulaion resuls on Raiop, n) for p = 0., p = 0.2, p = 0.1, and p = Generalized heurisic LR The consrucion mehod of he heurisic LR can be exended o he case when p is in a more general range. In [6], i was shown ha a sar is he opimal ree when p 1 1/) 1/, 1). Here, for a given value p 0, 1 1/) 1/ ) and a given number of leaves n, we show how o consruc an approximae ree GLRq, n) where q = 1 p). We found by simulaions ha he approximae ree performs very well. The idea of he consrucion is as follows. Firs, we find a good complee ernary ree; hen he roo of he GLRq, n) is forced o conain as many such good complee ernary rees as possible, wih one lefover ree. We will give some definiions before formally inroducing he heurisic. Definiion 15. When n =, we define a special group of n-progressions, SGn) = {T d n = {1,, 9,..., d, } d = 0, 1,..., 1}. For n =, we denoe c q) = min{cq, n T d n ) d = 0,..., 1} and le d q) be he value of d when c n q) is achieved. For example, when n = 27, SG27) = {{1, 27}, {1,, 27}, {1,, 9, 27}}. c q) = 27 min{1 q27, 1 q + 1 q27, 1 q + 1 q q27 }. 9 When q [1/) 1/, 1), cq, T 0 ) is 1 q, which is always larger han cq, T 1 ) = 1 q q ). We observe ha c q) can be calculaed recursively, c q) = 1 q + 1 c 1 q ), when > 1 and q [0.69, 1). There are wo base cases: one is c q) = 1 q ; he oher is c q) = 1 q when q 0, 1/) 1/ ), because he opimal ree is a sar when q 0, 0.69) [6]. From he recursive formula, we can ge he value of d q) = min{ log log q 1 )), 1}, where = log n. For a given q and n, we consruc an approximae ree GLRq, n) as follows: firs we calculae he value d q) = min{ log log q 1 )), 1}, where = log n, and consruc a complee ernary ree T q) of deph d q); he roo of GLRq, n) conains n subrees: n 1 of hem are T q) and one is GLRq, L 1 ), where n 1 = n d q) and L 1 = n mod d q). Take n = 0 and q = 0.9 as an example: d 0.9) = min{ log log )), log n 1} = 2. The good complee ernary ree has d = 9 leaves; he roo has four subrees, hree of which are complee ernary rees wih 9 leaves and one of which is a lefover subree GLR0.9, 2), which can be consruced recursively. The srucure of GLR0.9, 0) is shown in Fig. 4. We ran simulaions on he performance of GLR for various values of p and n. We redefine Raiop, n) = GLRp, n)/optp, n). Fig. 5 shows Raiop, n) as a funcion of n for p = 0., 0.2, 0.1, and Wihin he simulaion range, we see ha, for a fixed value p, he Raio curve oscillaes, bu ends o 0 as n ges larger. For p = 0.1, he maximum raio is below 1.02, when n is in he range [50, 250]. As p 0, boh he heurisic LR and he heurisic GLR mainain almos balanced

9 M. Li e al. / Theoreical Compuer Science ) ernary rees, bu hey deal wih he lefover leaves in wo differen ways: LR pus all he lefover leaves a he boom level of he complee ernary ree; GLR bundles all he lefover leaves as a lefover ree and pus i as a sibling of he complee ernary rees. In Fig. 5d), we compare he approximaion raios of LR and GLR when p = 0.01 and n = From he simulaion resuls, we see ha when n is large, GLR appears o perform beer. 7. Conclusions In his paper, we consider he group key managemen problem for broadcasing applicaions. In paricular, we focus on he case when membership changes are sparse. Under he assumpion ha every user has probabiliy p of being replaced by a new user during a bach rekeying period, he previously available algorihm requires On 4 ) ime o build he opimal key ree as p 0. We design a linear-ime heurisic LR o consruc an approximaely good key ree and analyze is performance as p 0. We prove ha LR produces a nearly 2-approximaion o he opimal key ree. Simulaion resuls show ha LR performs much beer han he heoreical bound we obain. We also design a refined heurisic LB whose approximaion raio is shown o be 1 + ɛ as p 0. A he end of his paper, we exended he heurisic LR o he algorihm GLR for he case when p is in a more general range p 0, 0.69) In [6], i was shown ha a sar is he opimal ree when p 0.07, 1).) Simulaion resuls show ha GLR performs very well as n increases. Some ineresing problems remain open. Alhough simulaion resuls show GLR performs very well for p 0, 0.69), we are no able o prove a consan approximaion raio for his range. In order o prove a consan bound, one needs o undersand beer he mahemaical srucure of he opimal n-progression S n ha achieves he value Fp, n) for arbirary p and n. References [1] S. Rafaeli, D. Huchison, A survey of key managemen for secure group communicaion, ACM Compuing Surveys 5 ) 200) [2] M.T. Goodrich, J.Z. Sun, R. Tamassia, Efficien ree-based revocaion in groups of low-sae devices, in: Proceedings of CRYPTO, [] C.K. Wong, M.G. Gouda, S.S. Lam, Secure group communicaions using key graphs, IEEE/ACM Transacions on Neworking 8 1) 200) 6 0. [4] X.S. Li, Y.R. Yang, M.G. Gouda, S.S. Lam, Bach re-keying for secure group communicaions, WWW10, Hong Kong, May 2 5, [5] F. Zhu, A. Chan, G. Noubir, Opimal ree srucure for key managemen of simulaneous join/leave in secure mulicas, in: Proceedings of MILCOM, 200. [6] R.L. Graham, M. Li, F.F. Yao, Opimal ree srucures for group key managemen wih bach updaes, SIAM Journal on Discree Mahemaics 21 2) 2007) [7] D. Wallner, E. Harder, R.C. Agee, Key managemen for mulicas: Issues and archiecures, RFC 2627, June 1999.

Longest Common Prefixes

Longest Common Prefixes Longes Common Prefixes The sandard ordering for srings is he lexicographical order. I is induced by an order over he alphabe. We will use he same symbols (,

More information

Some Ramsey results for the n-cube

Some Ramsey results for the n-cube Some Ramsey resuls for he n-cube Ron Graham Universiy of California, San Diego Jozsef Solymosi Universiy of Briish Columbia, Vancouver, Canada Absrac In his noe we esablish a Ramsey-ype resul for cerain

More information

T L. t=1. Proof of Lemma 1. Using the marginal cost accounting in Equation(4) and standard arguments. t )+Π RB. t )+K 1(Q RB

T L. t=1. Proof of Lemma 1. Using the marginal cost accounting in Equation(4) and standard arguments. t )+Π RB. t )+K 1(Q RB Elecronic Companion EC.1. Proofs of Technical Lemmas and Theorems LEMMA 1. Le C(RB) be he oal cos incurred by he RB policy. Then we have, T L E[C(RB)] 3 E[Z RB ]. (EC.1) Proof of Lemma 1. Using he marginal

More information

Final Spring 2007

Final Spring 2007 .615 Final Spring 7 Overview The purpose of he final exam is o calculae he MHD β limi in a high-bea oroidal okamak agains he dangerous n = 1 exernal ballooning-kink mode. Effecively, his corresponds o

More information

Random Walk with Anti-Correlated Steps

Random Walk with Anti-Correlated Steps Random Walk wih Ani-Correlaed Seps John Noga Dirk Wagner 2 Absrac We conjecure he expeced value of random walks wih ani-correlaed seps o be exacly. We suppor his conjecure wih 2 plausibiliy argumens and

More information

An introduction to the theory of SDDP algorithm

An introduction to the theory of SDDP algorithm An inroducion o he heory of SDDP algorihm V. Leclère (ENPC) Augus 1, 2014 V. Leclère Inroducion o SDDP Augus 1, 2014 1 / 21 Inroducion Large scale sochasic problem are hard o solve. Two ways of aacking

More information

1 Review of Zero-Sum Games

1 Review of Zero-Sum Games COS 5: heoreical Machine Learning Lecurer: Rob Schapire Lecure #23 Scribe: Eugene Brevdo April 30, 2008 Review of Zero-Sum Games Las ime we inroduced a mahemaical model for wo player zero-sum games. Any

More information

Notes for Lecture 17-18

Notes for Lecture 17-18 U.C. Berkeley CS278: Compuaional Complexiy Handou N7-8 Professor Luca Trevisan April 3-8, 2008 Noes for Lecure 7-8 In hese wo lecures we prove he firs half of he PCP Theorem, he Amplificaion Lemma, up

More information

Chapter 2. First Order Scalar Equations

Chapter 2. First Order Scalar Equations Chaper. Firs Order Scalar Equaions We sar our sudy of differenial equaions in he same way he pioneers in his field did. We show paricular echniques o solve paricular ypes of firs order differenial equaions.

More information

Vehicle Arrival Models : Headway

Vehicle Arrival Models : Headway Chaper 12 Vehicle Arrival Models : Headway 12.1 Inroducion Modelling arrival of vehicle a secion of road is an imporan sep in raffic flow modelling. I has imporan applicaion in raffic flow simulaion where

More information

Lecture 2-1 Kinematics in One Dimension Displacement, Velocity and Acceleration Everything in the world is moving. Nothing stays still.

Lecture 2-1 Kinematics in One Dimension Displacement, Velocity and Acceleration Everything in the world is moving. Nothing stays still. Lecure - Kinemaics in One Dimension Displacemen, Velociy and Acceleraion Everyhing in he world is moving. Nohing says sill. Moion occurs a all scales of he universe, saring from he moion of elecrons in

More information

23.2. Representing Periodic Functions by Fourier Series. Introduction. Prerequisites. Learning Outcomes

23.2. Representing Periodic Functions by Fourier Series. Introduction. Prerequisites. Learning Outcomes Represening Periodic Funcions by Fourier Series 3. Inroducion In his Secion we show how a periodic funcion can be expressed as a series of sines and cosines. We begin by obaining some sandard inegrals

More information

20. Applications of the Genetic-Drift Model

20. Applications of the Genetic-Drift Model 0. Applicaions of he Geneic-Drif Model 1) Deermining he probabiliy of forming any paricular combinaion of genoypes in he nex generaion: Example: If he parenal allele frequencies are p 0 = 0.35 and q 0

More information

On Boundedness of Q-Learning Iterates for Stochastic Shortest Path Problems

On Boundedness of Q-Learning Iterates for Stochastic Shortest Path Problems MATHEMATICS OF OPERATIONS RESEARCH Vol. 38, No. 2, May 2013, pp. 209 227 ISSN 0364-765X (prin) ISSN 1526-5471 (online) hp://dx.doi.org/10.1287/moor.1120.0562 2013 INFORMS On Boundedness of Q-Learning Ieraes

More information

Stability and Bifurcation in a Neural Network Model with Two Delays

Stability and Bifurcation in a Neural Network Model with Two Delays Inernaional Mahemaical Forum, Vol. 6, 11, no. 35, 175-1731 Sabiliy and Bifurcaion in a Neural Nework Model wih Two Delays GuangPing Hu and XiaoLing Li School of Mahemaics and Physics, Nanjing Universiy

More information

PENALIZED LEAST SQUARES AND PENALIZED LIKELIHOOD

PENALIZED LEAST SQUARES AND PENALIZED LIKELIHOOD PENALIZED LEAST SQUARES AND PENALIZED LIKELIHOOD HAN XIAO 1. Penalized Leas Squares Lasso solves he following opimizaion problem, ˆβ lasso = arg max β R p+1 1 N y i β 0 N x ij β j β j (1.1) for some 0.

More information

Matrix Versions of Some Refinements of the Arithmetic-Geometric Mean Inequality

Matrix Versions of Some Refinements of the Arithmetic-Geometric Mean Inequality Marix Versions of Some Refinemens of he Arihmeic-Geomeric Mean Inequaliy Bao Qi Feng and Andrew Tonge Absrac. We esablish marix versions of refinemens due o Alzer ], Carwrigh and Field 4], and Mercer 5]

More information

STATE-SPACE MODELLING. A mass balance across the tank gives:

STATE-SPACE MODELLING. A mass balance across the tank gives: B. Lennox and N.F. Thornhill, 9, Sae Space Modelling, IChemE Process Managemen and Conrol Subjec Group Newsleer STE-SPACE MODELLING Inroducion: Over he pas decade or so here has been an ever increasing

More information

Physics 235 Chapter 2. Chapter 2 Newtonian Mechanics Single Particle

Physics 235 Chapter 2. Chapter 2 Newtonian Mechanics Single Particle Chaper 2 Newonian Mechanics Single Paricle In his Chaper we will review wha Newon s laws of mechanics ell us abou he moion of a single paricle. Newon s laws are only valid in suiable reference frames,

More information

Article from. Predictive Analytics and Futurism. July 2016 Issue 13

Article from. Predictive Analytics and Futurism. July 2016 Issue 13 Aricle from Predicive Analyics and Fuurism July 6 Issue An Inroducion o Incremenal Learning By Qiang Wu and Dave Snell Machine learning provides useful ools for predicive analyics The ypical machine learning

More information

The Asymptotic Behavior of Nonoscillatory Solutions of Some Nonlinear Dynamic Equations on Time Scales

The Asymptotic Behavior of Nonoscillatory Solutions of Some Nonlinear Dynamic Equations on Time Scales Advances in Dynamical Sysems and Applicaions. ISSN 0973-5321 Volume 1 Number 1 (2006, pp. 103 112 c Research India Publicaions hp://www.ripublicaion.com/adsa.hm The Asympoic Behavior of Nonoscillaory Soluions

More information

Inventory Analysis and Management. Multi-Period Stochastic Models: Optimality of (s, S) Policy for K-Convex Objective Functions

Inventory Analysis and Management. Multi-Period Stochastic Models: Optimality of (s, S) Policy for K-Convex Objective Functions Muli-Period Sochasic Models: Opimali of (s, S) Polic for -Convex Objecive Funcions Consider a seing similar o he N-sage newsvendor problem excep ha now here is a fixed re-ordering cos (> 0) for each (re-)order.

More information

5. Stochastic processes (1)

5. Stochastic processes (1) Lec05.pp S-38.45 - Inroducion o Teleraffic Theory Spring 2005 Conens Basic conceps Poisson process 2 Sochasic processes () Consider some quaniy in a eleraffic (or any) sysem I ypically evolves in ime randomly

More information

Orthogonal Rational Functions, Associated Rational Functions And Functions Of The Second Kind

Orthogonal Rational Functions, Associated Rational Functions And Functions Of The Second Kind Proceedings of he World Congress on Engineering 2008 Vol II Orhogonal Raional Funcions, Associaed Raional Funcions And Funcions Of The Second Kind Karl Deckers and Adhemar Bulheel Absrac Consider he sequence

More information

Diebold, Chapter 7. Francis X. Diebold, Elements of Forecasting, 4th Edition (Mason, Ohio: Cengage Learning, 2006). Chapter 7. Characterizing Cycles

Diebold, Chapter 7. Francis X. Diebold, Elements of Forecasting, 4th Edition (Mason, Ohio: Cengage Learning, 2006). Chapter 7. Characterizing Cycles Diebold, Chaper 7 Francis X. Diebold, Elemens of Forecasing, 4h Ediion (Mason, Ohio: Cengage Learning, 006). Chaper 7. Characerizing Cycles Afer compleing his reading you should be able o: Define covariance

More information

Supplement for Stochastic Convex Optimization: Faster Local Growth Implies Faster Global Convergence

Supplement for Stochastic Convex Optimization: Faster Local Growth Implies Faster Global Convergence Supplemen for Sochasic Convex Opimizaion: Faser Local Growh Implies Faser Global Convergence Yi Xu Qihang Lin ianbao Yang Proof of heorem heorem Suppose Assumpion holds and F (w) obeys he LGC (6) Given

More information

3.1.3 INTRODUCTION TO DYNAMIC OPTIMIZATION: DISCRETE TIME PROBLEMS. A. The Hamiltonian and First-Order Conditions in a Finite Time Horizon

3.1.3 INTRODUCTION TO DYNAMIC OPTIMIZATION: DISCRETE TIME PROBLEMS. A. The Hamiltonian and First-Order Conditions in a Finite Time Horizon 3..3 INRODUCION O DYNAMIC OPIMIZAION: DISCREE IME PROBLEMS A. he Hamilonian and Firs-Order Condiions in a Finie ime Horizon Define a new funcion, he Hamilonian funcion, H. H he change in he oal value of

More information

11!Hí MATHEMATICS : ERDŐS AND ULAM PROC. N. A. S. of decomposiion, properly speaking) conradics he possibiliy of defining a counably addiive real-valu

11!Hí MATHEMATICS : ERDŐS AND ULAM PROC. N. A. S. of decomposiion, properly speaking) conradics he possibiliy of defining a counably addiive real-valu ON EQUATIONS WITH SETS AS UNKNOWNS BY PAUL ERDŐS AND S. ULAM DEPARTMENT OF MATHEMATICS, UNIVERSITY OF COLORADO, BOULDER Communicaed May 27, 1968 We shall presen here a number of resuls in se heory concerning

More information

Bernoulli numbers. Francesco Chiatti, Matteo Pintonello. December 5, 2016

Bernoulli numbers. Francesco Chiatti, Matteo Pintonello. December 5, 2016 UNIVERSITÁ DEGLI STUDI DI PADOVA, DIPARTIMENTO DI MATEMATICA TULLIO LEVI-CIVITA Bernoulli numbers Francesco Chiai, Maeo Pinonello December 5, 206 During las lessons we have proved he Las Ferma Theorem

More information

The Arcsine Distribution

The Arcsine Distribution The Arcsine Disribuion Chris H. Rycrof Ocober 6, 006 A common heme of he class has been ha he saisics of single walker are ofen very differen from hose of an ensemble of walkers. On he firs homework, we

More information

Single-Pass-Based Heuristic Algorithms for Group Flexible Flow-shop Scheduling Problems

Single-Pass-Based Heuristic Algorithms for Group Flexible Flow-shop Scheduling Problems Single-Pass-Based Heurisic Algorihms for Group Flexible Flow-shop Scheduling Problems PEI-YING HUANG, TZUNG-PEI HONG 2 and CHENG-YAN KAO, 3 Deparmen of Compuer Science and Informaion Engineering Naional

More information

Two Coupled Oscillators / Normal Modes

Two Coupled Oscillators / Normal Modes Lecure 3 Phys 3750 Two Coupled Oscillaors / Normal Modes Overview and Moivaion: Today we ake a small, bu significan, sep owards wave moion. We will no ye observe waves, bu his sep is imporan in is own

More information

Biol. 356 Lab 8. Mortality, Recruitment, and Migration Rates

Biol. 356 Lab 8. Mortality, Recruitment, and Migration Rates Biol. 356 Lab 8. Moraliy, Recruimen, and Migraion Raes (modified from Cox, 00, General Ecology Lab Manual, McGraw Hill) Las week we esimaed populaion size hrough several mehods. One assumpion of all hese

More information

Matlab and Python programming: how to get started

Matlab and Python programming: how to get started Malab and Pyhon programming: how o ge sared Equipping readers he skills o wrie programs o explore complex sysems and discover ineresing paerns from big daa is one of he main goals of his book. In his chaper,

More information

Pade and Laguerre Approximations Applied. to the Active Queue Management Model. of Internet Protocol

Pade and Laguerre Approximations Applied. to the Active Queue Management Model. of Internet Protocol Applied Mahemaical Sciences, Vol. 7, 013, no. 16, 663-673 HIKARI Ld, www.m-hikari.com hp://dx.doi.org/10.1988/ams.013.39499 Pade and Laguerre Approximaions Applied o he Acive Queue Managemen Model of Inerne

More information

Lecture Notes 2. The Hilbert Space Approach to Time Series

Lecture Notes 2. The Hilbert Space Approach to Time Series Time Series Seven N. Durlauf Universiy of Wisconsin. Basic ideas Lecure Noes. The Hilber Space Approach o Time Series The Hilber space framework provides a very powerful language for discussing he relaionship

More information

RC, RL and RLC circuits

RC, RL and RLC circuits Name Dae Time o Complee h m Parner Course/ Secion / Grade RC, RL and RLC circuis Inroducion In his experimen we will invesigae he behavior of circuis conaining combinaions of resisors, capaciors, and inducors.

More information

Lecture 2 October ε-approximation of 2-player zero-sum games

Lecture 2 October ε-approximation of 2-player zero-sum games Opimizaion II Winer 009/10 Lecurer: Khaled Elbassioni Lecure Ocober 19 1 ε-approximaion of -player zero-sum games In his lecure we give a randomized ficiious play algorihm for obaining an approximae soluion

More information

Designing Information Devices and Systems I Spring 2019 Lecture Notes Note 17

Designing Information Devices and Systems I Spring 2019 Lecture Notes Note 17 EES 16A Designing Informaion Devices and Sysems I Spring 019 Lecure Noes Noe 17 17.1 apaciive ouchscreen In he las noe, we saw ha a capacior consiss of wo pieces on conducive maerial separaed by a nonconducive

More information

Introduction to Probability and Statistics Slides 4 Chapter 4

Introduction to Probability and Statistics Slides 4 Chapter 4 Inroducion o Probabiliy and Saisics Slides 4 Chaper 4 Ammar M. Sarhan, asarhan@mahsa.dal.ca Deparmen of Mahemaics and Saisics, Dalhousie Universiy Fall Semeser 8 Dr. Ammar Sarhan Chaper 4 Coninuous Random

More information

Overview. COMP14112: Artificial Intelligence Fundamentals. Lecture 0 Very Brief Overview. Structure of this course

Overview. COMP14112: Artificial Intelligence Fundamentals. Lecture 0 Very Brief Overview. Structure of this course OMP: Arificial Inelligence Fundamenals Lecure 0 Very Brief Overview Lecurer: Email: Xiao-Jun Zeng x.zeng@mancheser.ac.uk Overview This course will focus mainly on probabilisic mehods in AI We shall presen

More information

Single and Double Pendulum Models

Single and Double Pendulum Models Single and Double Pendulum Models Mah 596 Projec Summary Spring 2016 Jarod Har 1 Overview Differen ypes of pendulums are used o model many phenomena in various disciplines. In paricular, single and double

More information

Simulation-Solving Dynamic Models ABE 5646 Week 2, Spring 2010

Simulation-Solving Dynamic Models ABE 5646 Week 2, Spring 2010 Simulaion-Solving Dynamic Models ABE 5646 Week 2, Spring 2010 Week Descripion Reading Maerial 2 Compuer Simulaion of Dynamic Models Finie Difference, coninuous saes, discree ime Simple Mehods Euler Trapezoid

More information

Application of a Stochastic-Fuzzy Approach to Modeling Optimal Discrete Time Dynamical Systems by Using Large Scale Data Processing

Application of a Stochastic-Fuzzy Approach to Modeling Optimal Discrete Time Dynamical Systems by Using Large Scale Data Processing Applicaion of a Sochasic-Fuzzy Approach o Modeling Opimal Discree Time Dynamical Sysems by Using Large Scale Daa Processing AA WALASZE-BABISZEWSA Deparmen of Compuer Engineering Opole Universiy of Technology

More information

) were both constant and we brought them from under the integral.

) were both constant and we brought them from under the integral. YIELD-PER-RECRUIT (coninued The yield-per-recrui model applies o a cohor, bu we saw in he Age Disribuions lecure ha he properies of a cohor do no apply in general o a collecion of cohors, which is wha

More information

Lecture 4 Notes (Little s Theorem)

Lecture 4 Notes (Little s Theorem) Lecure 4 Noes (Lile s Theorem) This lecure concerns one of he mos imporan (and simples) heorems in Queuing Theory, Lile s Theorem. More informaion can be found in he course book, Bersekas & Gallagher,

More information

A Hop Constrained Min-Sum Arborescence with Outage Costs

A Hop Constrained Min-Sum Arborescence with Outage Costs A Hop Consrained Min-Sum Arborescence wih Ouage Coss Rakesh Kawara Minnesoa Sae Universiy, Mankao, MN 56001 Email: Kawara@mnsu.edu Absrac The hop consrained min-sum arborescence wih ouage coss problem

More information

Rainbow saturation and graph capacities

Rainbow saturation and graph capacities Rainbow sauraion and graph capaciies Dániel Korándi Absrac The -colored rainbow sauraion number rsa (n, F ) is he minimum size of a -edge-colored graph on n verices ha conains no rainbow copy of F, bu

More information

Families with no matchings of size s

Families with no matchings of size s Families wih no machings of size s Peer Franl Andrey Kupavsii Absrac Le 2, s 2 be posiive inegers. Le be an n-elemen se, n s. Subses of 2 are called families. If F ( ), hen i is called - uniform. Wha is

More information

d 1 = c 1 b 2 - b 1 c 2 d 2 = c 1 b 3 - b 1 c 3

d 1 = c 1 b 2 - b 1 c 2 d 2 = c 1 b 3 - b 1 c 3 and d = c b - b c c d = c b - b c c This process is coninued unil he nh row has been compleed. The complee array of coefficiens is riangular. Noe ha in developing he array an enire row may be divided or

More information

Notes on Kalman Filtering

Notes on Kalman Filtering Noes on Kalman Filering Brian Borchers and Rick Aser November 7, Inroducion Daa Assimilaion is he problem of merging model predicions wih acual measuremens of a sysem o produce an opimal esimae of he curren

More information

Christos Papadimitriou & Luca Trevisan November 22, 2016

Christos Papadimitriou & Luca Trevisan November 22, 2016 U.C. Bereley CS170: Algorihms Handou LN-11-22 Chrisos Papadimiriou & Luca Trevisan November 22, 2016 Sreaming algorihms In his lecure and he nex one we sudy memory-efficien algorihms ha process a sream

More information

CMU-Q Lecture 3: Search algorithms: Informed. Teacher: Gianni A. Di Caro

CMU-Q Lecture 3: Search algorithms: Informed. Teacher: Gianni A. Di Caro CMU-Q 5-38 Lecure 3: Search algorihms: Informed Teacher: Gianni A. Di Caro UNINFORMED VS. INFORMED SEARCH Sraegy How desirable is o be in a cerain inermediae sae for he sake of (effecively) reaching a

More information

Guest Lectures for Dr. MacFarlane s EE3350 Part Deux

Guest Lectures for Dr. MacFarlane s EE3350 Part Deux Gues Lecures for Dr. MacFarlane s EE3350 Par Deux Michael Plane Mon., 08-30-2010 Wrie name in corner. Poin ou his is a review, so I will go faser. Remind hem o go lisen o online lecure abou geing an A

More information

Some Basic Information about M-S-D Systems

Some Basic Information about M-S-D Systems Some Basic Informaion abou M-S-D Sysems 1 Inroducion We wan o give some summary of he facs concerning unforced (homogeneous) and forced (non-homogeneous) models for linear oscillaors governed by second-order,

More information

Online Convex Optimization Example And Follow-The-Leader

Online Convex Optimization Example And Follow-The-Leader CSE599s, Spring 2014, Online Learning Lecure 2-04/03/2014 Online Convex Opimizaion Example And Follow-The-Leader Lecurer: Brendan McMahan Scribe: Sephen Joe Jonany 1 Review of Online Convex Opimizaion

More information

Computer-Aided Analysis of Electronic Circuits Course Notes 3

Computer-Aided Analysis of Electronic Circuits Course Notes 3 Gheorghe Asachi Technical Universiy of Iasi Faculy of Elecronics, Telecommunicaions and Informaion Technologies Compuer-Aided Analysis of Elecronic Circuis Course Noes 3 Bachelor: Telecommunicaion Technologies

More information

5.2. The Natural Logarithm. Solution

5.2. The Natural Logarithm. Solution 5.2 The Naural Logarihm The number e is an irraional number, similar in naure o π. Is non-erminaing, non-repeaing value is e 2.718 281 828 59. Like π, e also occurs frequenly in naural phenomena. In fac,

More information

Optimal Server Assignment in Multi-Server

Optimal Server Assignment in Multi-Server Opimal Server Assignmen in Muli-Server 1 Queueing Sysems wih Random Conneciviies Hassan Halabian, Suden Member, IEEE, Ioannis Lambadaris, Member, IEEE, arxiv:1112.1178v2 [mah.oc] 21 Jun 2013 Yannis Viniois,

More information

Expert Advice for Amateurs

Expert Advice for Amateurs Exper Advice for Amaeurs Ernes K. Lai Online Appendix - Exisence of Equilibria The analysis in his secion is performed under more general payoff funcions. Wihou aking an explici form, he payoffs of he

More information

Lab 10: RC, RL, and RLC Circuits

Lab 10: RC, RL, and RLC Circuits Lab 10: RC, RL, and RLC Circuis In his experimen, we will invesigae he behavior of circuis conaining combinaions of resisors, capaciors, and inducors. We will sudy he way volages and currens change in

More information

Course Notes for EE227C (Spring 2018): Convex Optimization and Approximation

Course Notes for EE227C (Spring 2018): Convex Optimization and Approximation Course Noes for EE7C Spring 018: Convex Opimizaion and Approximaion Insrucor: Moriz Hard Email: hard+ee7c@berkeley.edu Graduae Insrucor: Max Simchowiz Email: msimchow+ee7c@berkeley.edu Ocober 15, 018 3

More information

We just finished the Erdős-Stone Theorem, and ex(n, F ) (1 1/(χ(F ) 1)) ( n

We just finished the Erdős-Stone Theorem, and ex(n, F ) (1 1/(χ(F ) 1)) ( n Lecure 3 - Kövari-Sós-Turán Theorem Jacques Versraëe jacques@ucsd.edu We jus finished he Erdős-Sone Theorem, and ex(n, F ) ( /(χ(f ) )) ( n 2). So we have asympoics when χ(f ) 3 bu no when χ(f ) = 2 i.e.

More information

Two Popular Bayesian Estimators: Particle and Kalman Filters. McGill COMP 765 Sept 14 th, 2017

Two Popular Bayesian Estimators: Particle and Kalman Filters. McGill COMP 765 Sept 14 th, 2017 Two Popular Bayesian Esimaors: Paricle and Kalman Filers McGill COMP 765 Sep 14 h, 2017 1 1 1, dx x Bel x u x P x z P Recall: Bayes Filers,,,,,,, 1 1 1 1 u z u x P u z u x z P Bayes z = observaion u =

More information

Zürich. ETH Master Course: L Autonomous Mobile Robots Localization II

Zürich. ETH Master Course: L Autonomous Mobile Robots Localization II Roland Siegwar Margaria Chli Paul Furgale Marco Huer Marin Rufli Davide Scaramuzza ETH Maser Course: 151-0854-00L Auonomous Mobile Robos Localizaion II ACT and SEE For all do, (predicion updae / ACT),

More information

Section 4.4 Logarithmic Properties

Section 4.4 Logarithmic Properties Secion. Logarihmic Properies 59 Secion. Logarihmic Properies In he previous secion, we derived wo imporan properies of arihms, which allowed us o solve some asic eponenial and arihmic equaions. Properies

More information

ACE 562 Fall Lecture 5: The Simple Linear Regression Model: Sampling Properties of the Least Squares Estimators. by Professor Scott H.

ACE 562 Fall Lecture 5: The Simple Linear Regression Model: Sampling Properties of the Least Squares Estimators. by Professor Scott H. ACE 56 Fall 005 Lecure 5: he Simple Linear Regression Model: Sampling Properies of he Leas Squares Esimaors by Professor Sco H. Irwin Required Reading: Griffihs, Hill and Judge. "Inference in he Simple

More information

14 Autoregressive Moving Average Models

14 Autoregressive Moving Average Models 14 Auoregressive Moving Average Models In his chaper an imporan parameric family of saionary ime series is inroduced, he family of he auoregressive moving average, or ARMA, processes. For a large class

More information

KINEMATICS IN ONE DIMENSION

KINEMATICS IN ONE DIMENSION KINEMATICS IN ONE DIMENSION PREVIEW Kinemaics is he sudy of how hings move how far (disance and displacemen), how fas (speed and velociy), and how fas ha how fas changes (acceleraion). We say ha an objec

More information

Trees for Group Key Management with Batch Update

Trees for Group Key Management with Batch Update Trees for Group Key Management with Batch Update Ph. D. Defense May 30th, 2008 Nan Zang Outline Secure group key management overview Jumping sequence problem Properties of optimal jumping sequences Properties

More information

Exponential Weighted Moving Average (EWMA) Chart Under The Assumption of Moderateness And Its 3 Control Limits

Exponential Weighted Moving Average (EWMA) Chart Under The Assumption of Moderateness And Its 3 Control Limits DOI: 0.545/mjis.07.5009 Exponenial Weighed Moving Average (EWMA) Char Under The Assumpion of Moderaeness And Is 3 Conrol Limis KALPESH S TAILOR Assisan Professor, Deparmen of Saisics, M. K. Bhavnagar Universiy,

More information

2. Nonlinear Conservation Law Equations

2. Nonlinear Conservation Law Equations . Nonlinear Conservaion Law Equaions One of he clear lessons learned over recen years in sudying nonlinear parial differenial equaions is ha i is generally no wise o ry o aack a general class of nonlinear

More information

Most Probable Phase Portraits of Stochastic Differential Equations and Its Numerical Simulation

Most Probable Phase Portraits of Stochastic Differential Equations and Its Numerical Simulation Mos Probable Phase Porrais of Sochasic Differenial Equaions and Is Numerical Simulaion Bing Yang, Zhu Zeng and Ling Wang 3 School of Mahemaics and Saisics, Huazhong Universiy of Science and Technology,

More information

SOLUTIONS TO ECE 3084

SOLUTIONS TO ECE 3084 SOLUTIONS TO ECE 384 PROBLEM 2.. For each sysem below, specify wheher or no i is: (i) memoryless; (ii) causal; (iii) inverible; (iv) linear; (v) ime invarian; Explain your reasoning. If he propery is no

More information

Tom Heskes and Onno Zoeter. Presented by Mark Buller

Tom Heskes and Onno Zoeter. Presented by Mark Buller Tom Heskes and Onno Zoeer Presened by Mark Buller Dynamic Bayesian Neworks Direced graphical models of sochasic processes Represen hidden and observed variables wih differen dependencies Generalize Hidden

More information

Stochastic models and their distributions

Stochastic models and their distributions Sochasic models and heir disribuions Couning cusomers Suppose ha n cusomers arrive a a grocery a imes, say T 1,, T n, each of which akes any real number in he inerval (, ) equally likely The values T 1,,

More information

Chapter 7: Solving Trig Equations

Chapter 7: Solving Trig Equations Haberman MTH Secion I: The Trigonomeric Funcions Chaper 7: Solving Trig Equaions Le s sar by solving a couple of equaions ha involve he sine funcion EXAMPLE a: Solve he equaion sin( ) The inverse funcions

More information

SZG Macro 2011 Lecture 3: Dynamic Programming. SZG macro 2011 lecture 3 1

SZG Macro 2011 Lecture 3: Dynamic Programming. SZG macro 2011 lecture 3 1 SZG Macro 2011 Lecure 3: Dynamic Programming SZG macro 2011 lecure 3 1 Background Our previous discussion of opimal consumpion over ime and of opimal capial accumulaion sugges sudying he general decision

More information

Ensamble methods: Bagging and Boosting

Ensamble methods: Bagging and Boosting Lecure 21 Ensamble mehods: Bagging and Boosing Milos Hauskrech milos@cs.pi.edu 5329 Senno Square Ensemble mehods Mixure of expers Muliple base models (classifiers, regressors), each covers a differen par

More information

u(x) = e x 2 y + 2 ) Integrate and solve for x (1 + x)y + y = cos x Answer: Divide both sides by 1 + x and solve for y. y = x y + cos x

u(x) = e x 2 y + 2 ) Integrate and solve for x (1 + x)y + y = cos x Answer: Divide both sides by 1 + x and solve for y. y = x y + cos x . 1 Mah 211 Homework #3 February 2, 2001 2.4.3. y + (2/x)y = (cos x)/x 2 Answer: Compare y + (2/x) y = (cos x)/x 2 wih y = a(x)x + f(x)and noe ha a(x) = 2/x. Consequenly, an inegraing facor is found wih

More information

STA 114: Statistics. Notes 2. Statistical Models and the Likelihood Function

STA 114: Statistics. Notes 2. Statistical Models and the Likelihood Function STA 114: Saisics Noes 2. Saisical Models and he Likelihood Funcion Describing Daa & Saisical Models A physicis has a heory ha makes a precise predicion of wha s o be observed in daa. If he daa doesn mach

More information

Applying Genetic Algorithms for Inventory Lot-Sizing Problem with Supplier Selection under Storage Capacity Constraints

Applying Genetic Algorithms for Inventory Lot-Sizing Problem with Supplier Selection under Storage Capacity Constraints IJCSI Inernaional Journal of Compuer Science Issues, Vol 9, Issue 1, No 1, January 2012 wwwijcsiorg 18 Applying Geneic Algorihms for Invenory Lo-Sizing Problem wih Supplier Selecion under Sorage Capaciy

More information

Approximation Algorithms for Unique Games via Orthogonal Separators

Approximation Algorithms for Unique Games via Orthogonal Separators Approximaion Algorihms for Unique Games via Orhogonal Separaors Lecure noes by Konsanin Makarychev. Lecure noes are based on he papers [CMM06a, CMM06b, LM4]. Unique Games In hese lecure noes, we define

More information

Predator - Prey Model Trajectories and the nonlinear conservation law

Predator - Prey Model Trajectories and the nonlinear conservation law Predaor - Prey Model Trajecories and he nonlinear conservaion law James K. Peerson Deparmen of Biological Sciences and Deparmen of Mahemaical Sciences Clemson Universiy Ocober 28, 213 Ouline Drawing Trajecories

More information

Solutions from Chapter 9.1 and 9.2

Solutions from Chapter 9.1 and 9.2 Soluions from Chaper 9 and 92 Secion 9 Problem # This basically boils down o an exercise in he chain rule from calculus We are looking for soluions of he form: u( x) = f( k x c) where k x R 3 and k is

More information

RL Lecture 7: Eligibility Traces. R. S. Sutton and A. G. Barto: Reinforcement Learning: An Introduction 1

RL Lecture 7: Eligibility Traces. R. S. Sutton and A. G. Barto: Reinforcement Learning: An Introduction 1 RL Lecure 7: Eligibiliy Traces R. S. Suon and A. G. Baro: Reinforcemen Learning: An Inroducion 1 N-sep TD Predicion Idea: Look farher ino he fuure when you do TD backup (1, 2, 3,, n seps) R. S. Suon and

More information

Decentralized Stochastic Control with Partial History Sharing: A Common Information Approach

Decentralized Stochastic Control with Partial History Sharing: A Common Information Approach 1 Decenralized Sochasic Conrol wih Parial Hisory Sharing: A Common Informaion Approach Ashuosh Nayyar, Adiya Mahajan and Demoshenis Tenekezis arxiv:1209.1695v1 [cs.sy] 8 Sep 2012 Absrac A general model

More information

Math 333 Problem Set #2 Solution 14 February 2003

Math 333 Problem Set #2 Solution 14 February 2003 Mah 333 Problem Se #2 Soluion 14 February 2003 A1. Solve he iniial value problem dy dx = x2 + e 3x ; 2y 4 y(0) = 1. Soluion: This is separable; we wrie 2y 4 dy = x 2 + e x dx and inegrae o ge The iniial

More information

Morning Time: 1 hour 30 minutes Additional materials (enclosed):

Morning Time: 1 hour 30 minutes Additional materials (enclosed): ADVANCED GCE 78/0 MATHEMATICS (MEI) Differenial Equaions THURSDAY JANUARY 008 Morning Time: hour 30 minues Addiional maerials (enclosed): None Addiional maerials (required): Answer Bookle (8 pages) Graph

More information

Mathematical Theory and Modeling ISSN (Paper) ISSN (Online) Vol 3, No.3, 2013

Mathematical Theory and Modeling ISSN (Paper) ISSN (Online) Vol 3, No.3, 2013 Mahemaical Theory and Modeling ISSN -580 (Paper) ISSN 5-05 (Online) Vol, No., 0 www.iise.org The ffec of Inverse Transformaion on he Uni Mean and Consan Variance Assumpions of a Muliplicaive rror Model

More information

Martingales Stopping Time Processes

Martingales Stopping Time Processes IOSR Journal of Mahemaics (IOSR-JM) e-issn: 2278-5728, p-issn: 2319-765. Volume 11, Issue 1 Ver. II (Jan - Feb. 2015), PP 59-64 www.iosrjournals.org Maringales Sopping Time Processes I. Fulaan Deparmen

More information

Rapid Termination Evaluation for Recursive Subdivision of Bezier Curves

Rapid Termination Evaluation for Recursive Subdivision of Bezier Curves Rapid Terminaion Evaluaion for Recursive Subdivision of Bezier Curves Thomas F. Hain School of Compuer and Informaion Sciences, Universiy of Souh Alabama, Mobile, AL, U.S.A. Absrac Bézier curve flaening

More information

( ) a system of differential equations with continuous parametrization ( T = R + These look like, respectively:

( ) a system of differential equations with continuous parametrization ( T = R + These look like, respectively: XIII. DIFFERENCE AND DIFFERENTIAL EQUATIONS Ofen funcions, or a sysem of funcion, are paramerized in erms of some variable, usually denoed as and inerpreed as ime. The variable is wrien as a funcion of

More information

Ensamble methods: Boosting

Ensamble methods: Boosting Lecure 21 Ensamble mehods: Boosing Milos Hauskrech milos@cs.pi.edu 5329 Senno Square Schedule Final exam: April 18: 1:00-2:15pm, in-class Term projecs April 23 & April 25: a 1:00-2:30pm in CS seminar room

More information

Sections 2.2 & 2.3 Limit of a Function and Limit Laws

Sections 2.2 & 2.3 Limit of a Function and Limit Laws Mah 80 www.imeodare.com Secions. &. Limi of a Funcion and Limi Laws In secion. we saw how is arise when we wan o find he angen o a curve or he velociy of an objec. Now we urn our aenion o is in general

More information

EXERCISES FOR SECTION 1.5

EXERCISES FOR SECTION 1.5 1.5 Exisence and Uniqueness of Soluions 43 20. 1 v c 21. 1 v c 1 2 4 6 8 10 1 2 2 4 6 8 10 Graph of approximae soluion obained using Euler s mehod wih = 0.1. Graph of approximae soluion obained using Euler

More information

Learning to Take Concurrent Actions

Learning to Take Concurrent Actions Learning o Take Concurren Acions Khashayar Rohanimanesh Deparmen of Compuer Science Universiy of Massachuses Amhers, MA 0003 khash@cs.umass.edu Sridhar Mahadevan Deparmen of Compuer Science Universiy of

More information

6.2 Transforms of Derivatives and Integrals.

6.2 Transforms of Derivatives and Integrals. SEC. 6.2 Transforms of Derivaives and Inegrals. ODEs 2 3 33 39 23. Change of scale. If l( f ()) F(s) and c is any 33 45 APPLICATION OF s-shifting posiive consan, show ha l( f (c)) F(s>c)>c (Hin: In Probs.

More information

Modal identification of structures from roving input data by means of maximum likelihood estimation of the state space model

Modal identification of structures from roving input data by means of maximum likelihood estimation of the state space model Modal idenificaion of srucures from roving inpu daa by means of maximum likelihood esimaion of he sae space model J. Cara, J. Juan, E. Alarcón Absrac The usual way o perform a forced vibraion es is o fix

More information

Electrical and current self-induction

Electrical and current self-induction Elecrical and curren self-inducion F. F. Mende hp://fmnauka.narod.ru/works.hml mende_fedor@mail.ru Absrac The aricle considers he self-inducance of reacive elemens. Elecrical self-inducion To he laws of

More information