Assessing Attack Vulnerability in Networks with Uncertainty

Size: px
Start display at page:

Download "Assessing Attack Vulnerability in Networks with Uncertainty"

Transcription

1 Assessig Attack Vulerability i Networks with Ucertaity Thag N. Dih Dept. of CS, Virgiia Commowealth Uiversity Richmod, VA, USA, tdih@vcu.edu My T. Thai CISE Dept., Uiversity of Florida Gaiesville, FL, USA, 36, mythai@cise.ufl.edu Abstract A cosiderable amout of research effort has focused o developig metrics ad approaches to assess etwork vulerability. However, most of them eglect the etwork ucertaity arise due to various reasos such as mobility ad dyamics of the etwork, or oise itroduced i data collectio process. To this ed, we itroduce a framework to assess vulerability of etworks with ucertaity, modelig such etworks as probabilistic graphs. We adopt expected pairwise coectivity (EPC) as a measure to quatify global coectivity ad use it to formulate vulerability assessmet as a stochastic optimizatio problem. The objective is to idetify a few umber of critical odes whose removal miimizes EPC i the residual etwork. While solutios for stochastic optimizatio problems are ofte limited to small etworks, we preset a practical solutio that works for larger etworks. The key advatages of our solutio iclude ) the applicatio of a weighted averagig techique that avoids cosiderig all, expoetially may, possible realizatios of probabilistic graphs ad ) a Fully Polyomial Time Radomized Approximatio Scheme (FPRAS) to efficietly estimate the EPC with ay desired accuracy. Extesive experimets demostrate sigificat improvemet o performace of our solutio over other heuristic approaches. I. INTRODUCTION Networked systems such as commuicatio etworks, electrical grids, ad trasportatio etworks are vulerable to atural disasters ad targeted attacks. Eve failures of few vital odes or liks ca severely compromise the etwork s ability to meet its quality-of-service (QoS), if ot cause total etwork breakdow []. Moreover, there is a icreasig cocer over such critical systems as targets for (cyber) terrorist attacks []. To develop proactive resposes ad mitigate the risk, it is importat to assess etwork vulerability, i.e., to idetify those crucial odes ad liks, beforehad. Despite of may studies o assessig etwork vulerability, little is kow about assessmet of etworks with ucertaity. This ucertaity ca arise due to various reasos from mobility ad dyamics of etworks to data collectio process. Particularly, liks i techological etworks, e. g., the Iteret, wireless sesor etworks ad mobile opportuistic etworks, are frequetly subject to disruptios. Typical abstractio of etworks as static graphs [3], [4], that fails to capture this ucertaity, may lead to serious misjudgemet o etwork vulerability. I this paper, we propose a framework to assess vulerability of etworks with ucertaity. We model the etwork as a probabilistic graph G ad formulate the vulerability assessmet as a stochastic optimizatio problem. The goal Pr G i Pr G i (a) A probabilistic graph (b) Realizatios Fig. : A probabilistic graph with existig probabilities o edges. I (a), ad hoc heuristics which target odes of highest cetrality, e.g., degree ad betweeess, remove ode 3, leavig the residual etwork itact. I (b), miimizig EPC leads to the removal of ode 4. It results i 5% chace of breakig the residual etwork (i G ad G 4) ad disrupts effectively 55% etwork coectivity. is to idetify a small set of odes that removal miimizes expected value of etwork performace. We associate each edge i G with a probability of existece, represetig the fractio of time that the lik is i a workig state. Additioally, we treat G as a geerative model for determiistic graphs. Each such determiistic graph is a possible realizatio/sample of G ad is also associated with a probability of beig geerated. Our basic measure for etwork performace is pairwise coectivity, defied as the umber of coected ode pairs i the etwork. This measure has bee recetly adopted to accout for the impact of attacks i determiistic graphs [3], [4], [5], [6]. It is favored for the strog discrimiatio i quatifyig the etwork coectivity level, eve for discoected etworks. Give a probabilistic graph G, we defie k-pcnd as the problem of fidig k odes that removal miimizes the expected pairwise coectivity (EPC), over all possible realizatios of G. The advatage of our assessmet framework over existig approaches is illustrated i Fig.. If oe ode is to remove from the graph accordig to either degree cetrality or betweeess cetrality, ode 3 will be removed. As show i Fig. a, the residual etwork remais coected i this case. However, Fig. b shows a more destructive attack o ode 4 through miimizig EPC. Sice most liks i the etwork have existig probabilities oe, except for two liks (, 3) ad (3, 5), we have four possible realizatios of the etwork, amed from G to G 4. Removig ode 4 will ot oly degrade EPC by

2 55% but also result i a 5% chace of breakig the residual etwork, as i the cases of G ad G 4. More sophisticated assessmet methods [4], [5] also fail o this simple example, idicatig the importace of cosiderig etwork ucertaity. Aother advatage of our approach is that we are able to costruct a efficiet Fully Polyomial-Time Radomized Approximatio Scheme (FPRAS) to compute EPC. Our FPRAS does ot cosider all, expoetially may, possible realizatios of G, yet it computes EPC with guarateed accuracy. As exact computatio of EPC is #P-complete [7], a FPRAS provides the best theoretical result. Such a result is ot kow for ay other measures for probabilistic graphs except all-termialreliability which admits the oly kow FPRAS i 995 [8]. Ulike the first FPRAS [8] which is oly practical for small etworks with equal edge probabilities, our FPRAS is scalable for large etworks with heterogeeous edge probabilities. Last but ot least, stochastic optimizatio problems are extremely difficult to solve. Commo techiques for stochastic programmig problems such as Beder s decompositio [9] ad Sample Average Approximatio [] ofte do ot scale beyod etworks with few dozes of odes. Thus it is critical that we desig a efficiet solutio for the problem. We summarize our cotributios as follows: We itroduce a framework to assess attack vulerability i etworks with ucertaity, formulatig it as stochastic optimizatio problems over probabilistic graphs. Besides EPC, the framework ca be itegrated with may other reliability ad performace measures such as size of largest compoets ad average maximum flow betwee ode pairs []. We formulate k-pcnd problem to assess etwork vulerability ad preset a practical solutio for the problem which utilizes a weighted averagig techique to avoid cosiderig all, expoetially may, possible realizatios of probabilistic graphs. We propose a FPRAS for computig EPC. The FPRAS is ot oly of theoretical iterest but also practical for large etworks. Extedig techiques i our FPRAS to other reliability problems, e.g., the two-termialreliability ad k-termial-reliability [8], is material for future work. We show sigificat performace improvemet of our algorithm over competitor heuristics via experimets. The experimets also reveal that the vulerability assessmet based o determiistic etwork aalysis is too optimistic i real scearios, as it greatly overestimates how resiliet etwork systems are. Orgaizatio. We summarize related work i Sectio II ad itroduce models, otatios, ad defiitio of the k-pcnd problem i Sectio III. Assumig the presece of a efficiet oracle to estimate EPC, we preset our algorithm for k-pcnd i Sectio IV. We preset our FPRAS to estimate EPC, the fial piece i our solutio, i Sectio V. We give simulatio results o real-world traces to show efficacy of our algorithm ad the isights i Sectio VI. II. RELATED WORK To our best kowledge, we are the first to study attack vulerability assessmet for ucertai etworks, optimizig directly a reliability measure desiged for such etworks. However, attack vulerability i determiistic etworks ad reliability (i the cotext of radom failures) for etworks with ucertaity have motivated may studies i commuicatio ad theoretical commuities. A. Vulerability Assessmet i Determiistic Networks. The fuctio ad performace of etworks rely o their resiliece, defied as the ability to cotiue fuctioig uder perturbatio. To measure robustess ad resiliet, prior works proposed to moitor differet measures icludig the diameter, size of the largest coected compoet, [], coectivity based o miimum ode/edge cut, algebraic coectivity, ad spectral radius [], [3], [5]. All these measures capture, i priciple, aspects of etwork coectivity. May metrics ad approaches have bee proposed to accout for etwork robustess ad vulerability [], [3], [6]. While each of these measures has its ow emphasis ad ratioality, they ofte come with several shortcomigs that prevet them from capturig desired characteristics of etwork coectivity ad resiliece. For example, measures based o shortest path are rather sesitive to small chages (e.g. removig edges or odes); algebraic coectivity ad diameter are ot meaigful for discoected graphs (all discoected graphs have the same values); umber of coected compoets ad compoet sizes, arguably, do ot fully reflect level of etwork coectivity. Pairwise coectivity, defied as the umber of ode pairs that remai coected, has bee proposed as a effective measure for etwork coectivity [3], [4], [5], [4], [6]. Arulselva et al. defied i [4] the Critical Node Detectio (CND) problem, the determiistic versio of k-pcnd. The problem seeks k odes that removal miimize the pairwise coectivity i the residual etwork. We proposed β-disruptor framework to assess attack vulerability i terms of pairwise coectivity [5], [4]. We preseted O ( log.5 ) bicriteria approximatio algorithms for assessig edge vulerability, ad a O (log log log ) bicriteria approximatio algorithm for the vertex versio of β-disruptor. Whe both odes ad liks i the etwork are subjected to attacks, we provide a O( log ) bicriteria approximatio algorithm [5] that immediately improve the results i [5], [4]. B. Reliability of Networks with Radom Failures. A sigificat amout of works has bee devoted for reliability of etworks whe their elemets are subjected to radom failures [6], [7], [7]. The well studied reliability assessmet framework is to calculate the probability that commuicatio ca be established amog a set of odes whe each ode ad/or lik ca fail idepedetly with some probability. The two-termial-reliability, with two special odes called source s ad destiatio t, cocers the probability that there exists a path betwee s ad t. More geeral cases ivolve k-termial reliability ad all-termial-reliability. The etwork reliability

3 problems were proved to be i #P-hard class, a super class of NP-hard problems [8]. Karger itroduced the first FPRAS for all-termial-reliability problem [8]. To our best kowledge, it is the oly kow FPRAS for etwork reliability problems. I [7], [9], Colbour itroduced etwork resiliece, defied as the average two-termial-reliability betwee all ode pairs. The exact computatio of etwork resiliece was proved to be #P-hard eve i plaar graphs. Ami et al. [] proposed pair-coected reliability, the expected umber of coected ode pairs. We ote that our measure EPC is the same as Pair-coected reliability ad both ca be obtaied by multiplyig etwork resiliece by ( ) V. We use the ame expected pairwise coectivity to be cosistet with amig of pairwise coectivity i [3], [4], [5]. Recetly, Neumayer et al. [] proposed a polyomial-time algorithm to compute etwork resiliece, which they referred to as average twotermial reliability (ATTR) [], [], for geometric etworks whe the disaster takes a special form of a straight lie. However, efficiet methods to compute etwork resiliece i geeral case is still a ope questio til this paper. We ote that the reliability literature, however, does ot cosider targeted attacks, which is the mai subject of this paper. III. MODEL AND DEFINITIONS I this sectio, we preset the probabilistic etwork model, ad the ecessary otatios to formulate the vulerability assessmet problem. A. Probabilistic Network Model We abstract a etwork with ucertaity as a probabilistic graph G = (V, E, P ) where vertices i V correspods to the set of odes; edges i E correspods to the set of liks i the etwork; ad P that maps each edge (u, v) E to a real umber i p uv [, ] that represets the probability that edge (u, v) exists. For each (u, v) / E, we have p uv =. A example of probabilistic graphs is Erdos-Reyi radom graphs [3] i which all edge probabilities are the same ad equal p. For clarity, we cosider oly udirected etworks ad assume idepedece amog edges. However our proposed solutio also applies i priciple to directed graphs or graphs with edge correlatios as log as expected values of edges ca be computed. A probabilistic graph G ca be see as a geerative model for determiistic graphs. A determiistic graph G = (V, E s ) is geerated from G by selectig each edge (u, v) E, idepedetly, with probability p uv. We refer to G as a realizatio or a sample of G ad write G G. The probability that G is geerated from G is Pr[G] = p e ( p e ). s \E s Let m = E, there are W = m possible realizatios of G. We umber those realizatios as G = (V, E ), G = (V, E ),..., G W = (V, E W ), where E, E,..., E W are all possible subsets of E. B. Expected pairwise coectivity Our mai measure for the etwork reliability is expected pairwise coectivity (EPC), which is the expected umber of coected pairs i the etwork. Formally, deote by P(G) the umber of coected pairs or pairwise coectivity of a determiistic graph G. The expected pairwise coectivity (EPC) of G is defied as EPC(G) = E [P(G)] = G G Pr[G]P(G). EPC has a tight coectio to two-termial reliability as stated i the followig lemma. Lemma : [7] Let REL u,v (G) deote the two-termialreliability betwee ode u ad v i G, i.e. the probability that v is reachable from u. We have EPC(G) = REL u,v (G). () u,v V ;u v Thus EPC ca be computed as the total of two-termialreliability betwee all ode pairs. However, this approach is problematic due to the facts that exact computatio for twotermial-reliability is NP-hard, ad that eve we apply existig heuristics to approximate two-termial-reliability, computig EPC requires a large umber, ( ) V, of calls to such heuristics. Istead, EPC ca be computed efficietly as show i Sectio V. This is a importat advatage of EPC over other reliability measures ad the reaso for the adoptio of EPC. C. Vulerability Assessmet i Probabilistic Networks We formulate vulerability assessmet as the followig stochastic optimizatio problem. Probabilistic Critical Nodes Detectio (k-pcnd). Give a probabilistic etwork G = (V, E, p) ad a iteger k, fid a k odes subset S V that removal miimizes EPC i the residual etwork. Whe all edge probabilities are oe, we obtai the CND problem i [4]. Sice the CND problem is NP-hard, k-pcnd, geeralizig CND, is also NP-hard. IV. VULNERABILITY ASSESSMENT IN PROBABILISTIC NETWORKS I this sectio, we ivestigate the Probabilistic Critical Nodes Problem (k-pcnd). We formulate the problem as a two-stage stochastic program i Subsectio IV-A; ad devise efficiet approaches to overcome the difficulty of havig a expoetial umber of costraits i the mathematical formulatio i Subsectio IV-B. Our solutio assume the presece of a efficiet oracle to compute EPC, which we preset later i Sectio V. A. Two-stage Stochastic Liear Program Stochastic programmig has bee a commo approach for optimizatio uder ucertaity whe the probability distributio govers the data is give. A comprehesive itroductio to stochastic programmig ca be foud i referece [5]. Give a probabilistic graph G = (V, E, p) ad a iteger < k <, we first use iteger variables s i to represet whether or ot ode i is removed, i.e, s i = if ode i is

4 removed, ad, otherwise. Here = V is the umber of odes ad we assume odes are umbered from to. We impose o s the costrait i= s i k to guaratee o more tha k odes are removed. Variables s are kow as first stage variables. The values of s are to be decided before the actual realizatio of the ucertai parameters i G. We associate with each ode pair (u, v) a radom Berouli variable ξ uv satisfyig Pr[ξ uv = ] = p uv ad Pr[ξ uv = ] = p uv. For each realizatio of G, the values of ξ uv are revealed to be either or ad we ca compute the pairwise coectivity i the residual graph after removig k odes idicated by s. To do so, we defie iteger variables x ij to be the discoectivity betwee a ode pair i ad j i the residual etwork, i.e., x ij = if i ad j are still coected ad, otherwise. Pairwise coectivity i the residual graph ca be computed usig a secod stage iteger programmig, deoted by P (s, x, ξ) as follows. P (s, x, ξ) = mi i<j( x ij ) () s. t. x ij s i + s j + ξ ij, (i, j) E, (3) x ij + x jk x ik, (i, j) E, k =.. (4) s i {, }, x ij {, } (5) This secod stage programmig formulatio is essetially the same with the formulatio for the CND problem i [4] (except for the cardiality costrait o s). Ideed, we adopt the improved formulatio of CND i [4]. This formulatio reduces the umber of costraits from θ( 3 ) to O(m) ad shorte the solvig time substatially. The two-stage stochastic liear formulatio for the k-pcnd problem is as follows. mi E [P (s, x, ξ)] (6) s {,} s. t. s i k (7) i= where P (s, x, ξ) is give i ()-(5) (8) The objective is to miimize the expected coectivity i the residual etwork E [P (s, x, ξ)], where P (s, x, ξ) is the optimal value of the secod-stage problem. This stochastic programmig problem is, however, ot yet ready to be solved with liear algebra solver. Discretizatio. To solve a two-stage stochastic problem, oe ofte eed to discretize the problem ito a sigle (very large) liear programmig problem. That is we eed to cosider all possible realizatios G l G ad their probability masses Pr[G l ]. Deote by {ξ l } ij the adjacecy matrix of the realizatio G l = (V, E l ), i.e., ξij l =, if (i, j) El, ad, otherwise. Sice the objective ivolves oly the expected cost of the secod stage variables x ij, the two-stage stochastic program ca be discretized ito a mixed iteger programmig, deoted by MIP F as follows. mi W ( x l ij) (9) Pr[G l ] l= i<j s. t. s i k () i= x l ij s i + s j + ξ l ij, (i, j) E, l =..W () x l ij + x l jk x l ik, (i, j) E, k =.., l =..W () x l ij = x l ji, i, j =.., l =..W (3) s {, }, x l [, ], l =..W (4) The major challege i solvig this discretized form is that there is a expoetial umber of variables ad costraits. Thus, solvig MIP F is itractable eve for very small istaces of G. To overcome this difficulty, we preset i ext subsectio a compact relaxatio of MIP F. Solvig this polyomial size relaxatio leads to high quality solutios for k-pcnd problem, as we will show i the experimetal sectio. B. Algorithm We preset our solutio, amed, for the stochastic optimizatio problem i Algorithm. First, the algorithm costructs a liear relaxatio of the expoetial size formula MIP F ad select k vertices via a iterative roudig procedure. The result is a subset D of cardiality k. The algorithm follows by a local search procedure that improves D via swappig vertices. A vertex u D ad a vertex v / D are swapped places if doig so reduces the EPC. The key of the local search is to compute EPC quickly ad accurately. This is doe with the CSP algorithm (preseted later i Algorithm ). The local search stops whe o more swaps ca reduce the EPC. The relaxatio of MIP F is costructed by applyig a weighted-averagig of all costraits i MIP F. Costraits ivolvig the realizatio G l are give weights Pr[G l ]. Thus costraits () are reduced to a sigle costrait x ij s i + s j + Pr[G l ]ξij, l G l G which ca be further simplified ito x ij s i +s j + p ij. The other costraits ca be averaged i the same way, givig us the followig relaxatio of MIP F. mi ( x ij ) (5) s. t. i<j s i k (6) i= x ij s i + s j + p ij, (i, j) E (7) x ij + x jk x ik, (i, j) E, k =.. (8) s i {, }, x ij [, ], i, j =.. (9) We shall refer to this relaxatio of MIP F as MIP R. Note that the o-itegrality of x ij is essetial for the above relaxatio, deoted by MIP R. If we restrict x ij to {, } the costrait x ij s i +s j ξ ij is equivalet to x ij s i +s j +.

5 The the costrait holds trivially for ay values of x ad s. Thus the iformatio ecoded i the edge probabilities is ot itegerated i the formulatio. Algorithm. Roudig the Expected Graph Algorithm () ) Obtai a LP relaxatio of MIP R with the relaxed costraits s [, ]. ) Iitialize the set of selected odes D =. 3) Repeat k times the followig steps Solve the LP relaxatio Select u = arg max i V \D s i. Add u to D ad set s u 4) Repeat 5) For each pair (u, v) D (V \ D) 6) Estimate the EPC after removig D {u} + {v} usig CSP 7) Update D = D {u} + {v}, if the ew EPC is lower 8) Util o possible update 9) Output D. Lower-boud. Oe of the ice feature of MIP R is that its optimal objective provides a lower-boud o the optimal objective of MIP F. This provides a useful tool to assess the quality of proposed algorithms, especially whe fidig the optimal solutios of MIP F is likely itractable. We prove the objective lower-boud i the followig lemma. Lemma : The optimal objective value of the MIP R is a lower-boud o the optimal objective value of MIP F. Proof: To show that the the objective of the MIP R is a lower-boud o that of the MIP F, we costruct a feasible solutio ( s, x) of MIP R that gives a objective equal to the optimal objective of MIP F. Let ( ŝ, ˆx,..., ˆx ) W ( be a optimal solutio of the MIP F. Costruct a solutio s = ŝ, x = W l= Pr[Gl ]ˆx ). l The objective value of MIP R give by that solutio is x ij ) = i<j( W ( Pr[G l ]ˆx l ij) i<j l= = W Pr[G l ]( ˆx l ij) i<j which is exactly the optimal objective of MIP F. The last equality holds because the probabilities Pr[G l ] add up to oe. The rest is to show that ( s, x) is a feasible solutio of MIP R. Clearly, s satisfy (6) ad the itegral costraits. Also sice x is a covex combiatio of ˆx l, l =..W with the masses Pr[G l ], x satisfy the costraits (8), (7),& (9) as they ca be iferred from the same covex combiatio of the costraits from () to (4). We ote that due to the high similarity i programmig formulatios of critical elemets detectio problems, our algorithm ca be easily modified to solve extesios of other vulerability assessmet problems to etworks with ucertaity. Examples iclude the Critical Edge Detectio [6], β-vertex disruptor, ad β-edge disruptor [5], [6]. l= V. COMPUTING EPC This sectio focuses o efficiet methods to compute EPC of probabilistic graph, the fial but importat piece of the algorithm. Sice it is itractable to compute the exact value of EPC [7], we preset efficiet methods to approximate EPC with ay desired accuracy. A. Compoet Samplig Procedure to Approximate EPC We develop a Mote Carlo method to approximate the EPC withi a arbitrary small error with a high probability. We also reveal why the aive Mote Carlo method caot guaratee a polyomial time complexity. Give a pair of ɛ, δ >, our Mote Carlo method returs a estimatio of EPC(G) accurate to withi a relative error of ɛ with a probability at least δ. Mathematically, our proposed method is a (ɛ, δ)-approximatio of EPC, which is defied as follows. Defiitio ( (ɛ, δ)-approximatio): A fuctio ˆF (G) is a (ɛ, δ)-approximatio for the expected pairwise coectivity EPC(G) if [ Pr ( ɛ)epc(g) ˆF ] (G) ( + ɛ)epc(g) > δ. A (ɛ, δ)-approximatio is called a fully polyomial radomized approximatio scheme (FPRAS) if its ruig time is bouded by a polyomial i terms of /ɛ, log(/δ), ad the iput size. I geeral, a FPRAS is the best theoretical result oe ca hope for a #P-hard computatioal problem. We preset our Compoet Samplig Procedure (CSP) with two importat advatages over the aive Mote Carlo method. First, it has a polyomial time complexity ad is, thus, a FPRAS for the EPC(G) problem. Secod, it has a smaller average time complexity, ad is up to times faster tha aive Mote Carlo methods. CSP is summarized i Algorithm. The algorithm computes the sum of edge probabilities P E = p e. If P E is sufficietly small (at most ɛ ), the algorithm returs P E as a ubiased estimator of EPC(G). Otherwise, it performs a importace samplig method to estimate EPC(G) i steps 4 to 6. I the importace samplig method, we select a ode u V uiformly ad perform a Bread-First Search procedure from u, util reachig all odes i the coected compoet that cotais u. The algorithm the computes the average of the size of the compoet that cotais u less oe, ad multiply the result by to obtai a ubiased estimator E. Oe advatage of CSP over direct Mote-Carlo approaches is that it avoids geeratig too may graph samples whe the EPC is predicted to be small. This is the key to guaratee that the algorithm is polyomial-time. Further, the algorithm does ot geerate the whole sample graph at oce, but oly reveal the availability of edges alog the Bread-First Search procedure. This characteristic substatially reduces CSP s average ruig time, as aalyzed later i Theorem 3.

6 Algorithm. (ɛ, δ) Compoet Samplig Procedure to Approximate EPC(G) ) Let P E = pe ) if P E < ɛ the 3) retur E = P E. 4) C. 5) for i = to N(ɛ, δ) do Select a ode u V uiformly. Start a Breath-First Search from u. For each ecoutered edge (v, w), flip a coi of bias p vw to determie its availability. Let S i be the umber of visited odes, icludig ode u. C = C + (S i ). 6) Retur E = C N B. Correctess as a ubiased estimator of EPC(G). We determie whether or ot the value EPC is too small based o the value of P E = p e. This is based o a observatio that EPC is sadwiched betwee two fuctios of P E, as show i the followig propositio. Propositio : Let G = (V, E, p) be a probabilistic graph ad P E = p e, the followig iequality holds ( P E EPC(G) + P ) m E. () m The bouds i Propositio are asymptotic tight i the sese that there are arbitrary large graphs i which the bouds are oly differet from the actual values of EPC(G) by a factor of two. For example, cosider G as a star graph of size that cosists of oe ceter vertex ad leaves. All edges are assiged the same probability /( ). Oe ca verify that the lower-boud, EPC(G), ad the upper boud are, 3 ( ), ad ( + ) < e, respectively. Further, we state several iequalities eeded for (ɛ, δ)- approximatio proof i the followig propositio. Propositio : Let G = (V, E, p) be a probabilistic graph ad q i be the probability that G has exactly i edges, for i =.. E. If P E = p e < /, the followig iequalities hold P E q exp( P E ), () P E q P E q q, () P E m q k PE. (3) k= For the sake of completeess, we preset the proofs of Propositio ad i the Appedix. We ow derive N(ɛ, δ), the umber of ecessary samples to be draw usig the followig Geeralized Zero-Oe Estimator Theorem itroduced by Dagum et al. [7]. Theorem : (Geeralized Zero-Oe Estimator [7]) Let X, X,..., X N be idepedet idetically distributed radom variables takig values i [, ], with mea µ >. If < ɛ < ad N 4(e ) l(/σ)/(ɛ µ), where e.78 is Euler s umber, the [ Pr ( ɛ)µ N ] N X i ( + ɛ)µ > δ. i= The required umber of samples to obtai a (ɛ, δ) approximatio is N(ɛ, δ) = 4(e ) l ( ) σ ɛ EPC(G), as proved i the followig lemma. Lemma 3: If N(ɛ, δ) 4(e ) l ( ) σ ɛ EPC(G), the E, the output of CSP, is a (ɛ, δ)-approximatio for EPC(G). Proof: We cosider two cases of P E. Case P E < ɛ : CSP returs P E (step ). We show that P E is ideed a (ɛ, δ)-approximatio for EPC(G) by provig the followig iequalities. P E EPC(G) ( + ɛ)p E. From Lemma, we already have P E EPC(G). Thus we oly eed to show EPC(G) ( + ɛ)p E. Deote by q k, k =..m, the probability that G has exactly k edges. Sice i ay (determiistic) graph with m edges ad vertices, the pairwise coectivity caot exceeds ( mi{m,} ), we obtai the followig iequality. EPC(G) ( ) q + ( ) q + ( ) m q k (4) Apply iequalities (), ad (3) p E EPC(G) q + ( p E ) P E. Usig iequality (), we arrive P E k= EPC(G) exp( P E ) + ( P E ) P E. Sice P E ɛ < /, we apply the iequality exp( x) x + x, x (, ) to yield EPC(G) ( + P E + ( ) PE ) PE ( + ɛ)p E. (5) This completes the proof of P E beig (ɛ, δ)-approximatio of EPC(G) whe P E < ɛ. Case P E ɛ : The importace samplig is carried out i steps 4 to 6 i Algorithm. Withi the loop i Step 5, we ca compute S i with the followig equivalet procedure: ) Draw a sample graph G i ; ad ) Select a ode u i G i uiformly ad compute S i as the size of coected compoet that cotais u. Assume that there are t coected compoets with sizes s, s,..., s t i G i. We have E[S i G = G i ] = k i= s i(s i ) k i= s i = P(Gi ).

7 EPC Time(secods) Hece (a) Erdos-Reyi etwork EPC (b) Barabasi-Albert Network EPC (c) Small-world Network EPC (d) US backboe etwork Fig. : Comparig performace of the algorithms o differet etwork topologies ad edge probabilities Time(secods) (a) Erdos-Reyi etwork (b) Barabasi-Albert Network (c) Small-world Network (d) US backboe etwork Fig. 3: Comparig ruig time of the algorithms (y-axis) with differet edge probabilities (x-axis). E[S i ] = EPC(G)/. (6) By applyig Theorem to i.i.d. radom variables Y i = (S i )/( ) with mea µ = EPC(G)/ ( ), it follows that E is a (ɛ, δ)-approximatio of EPC(G). C. Time Complexity Aalysis Lemma 4: CSP has a time complexity O(m 4 ɛ 3 ). Proof: If P E < ɛ, the algorithm takes a O(m) time, as we oly eed to compute P E. Otherwise, the algorithm performs N(ɛ, δ) times the BFS algorithm i step 5. Sice the BFS algorithm takes a time at most O(m + ), the total time take by Algorithm is upper bouded by θ(m + )4(exp ) l δ ( ) ɛ EPC(G) By Propositio, EPC(G) P E ɛ. Thus the worstcase time complexity is at most O(m 4 ɛ 3 ). Lemmas 3 ad 4 immediately lead to our mai result for approximatig EPC, stated i the followig theorem. Theorem (Mai theorem): CSP is a FPRAS for the EPC computatio problem that outputs a (ɛ, δ)-approximatio of EPC i a O(m 4 ɛ 3 ) time. CSP ot oly has a polyomial ruig time, it s also faster tha aive Mote Carlo methods. I geeral, the time eeded for the BFS procedure i CSP is ofte less tha the time to geerate a sample graph. The reaso is that the BFS procedure oly eeds to be aware about the surroudig of the selected vertex u, while geeratig a graph sample might ivolve all edges ad vertices i the graph. To formally prove this observatio, we give the expected ruig time of CSP i the followig theorem. ( Theorem 3: CSP has a expected time complexity O ɛ mi{ + m EPC(G), m EPC(G) ). } The proof of this theorem ca be foud i the Appedix. Time(secods) Time(secods) VI. EXPERIMENTS We demostrate through our experimets the efficiecy of our proposed algorithm ad the eed of ew assessmet methods for etworks with ucertaity. A. Experimet Setup Dataset. We aalyze the performace our proposed algorithm through experimets o differet etwork models ad a real commuicatio etwork, as described below. Erdos-Reyi: A radom graph of vertices ad edges followig the Erdos-Reyi model [3]. Barabasi-Albert: A radom graph of vertices ad edges. The graph follows power-law model usig preferetial attachmet mechaism [8]. Watts Strogatz: A radom graph that is geerated from the small-world model [9] with the dimesio of the lattice ad the rewirig probability.3 [9]. US Backboe etwork: The US backboe cablig etwork of XO compay [3] with 78 odes ad 9 liks. Compared Methods. We compare the performace of, Algorithm, with the followig methods : Sample Average Approximatio method [], a commo techique to solve stochastic optimizatio problem. The solutio is furthered optimize usig the same local search procedure i. The umber of samples to optimize is T = 3., a greedy algorithm that removes the odes with the highest betweeess cetrality values., aother heuristic that removes the edge with the highest values. The dampig factor is.85. The umber of samples draw i the local search procedure i both ad are. Durig our experimets, we observe that the local search procedure are quite isesitive to the umber of samplig times. The fial EPC i each etwork is, however, measured by settig the umber of sample times to, to guaratee high accurate estimatio of EPC.

8 Eviromet. All algorithms are implemeted i C++ ad compiled with GCC 4.4 compiler o a 64 bit Widow machie with a i7 3.4Ghz processor ad 6 GB memory. The mathematical optimizatio package to solve liear programmig formulatio is GUROBI 5.5. B. Experimetal results. The solutio quality, i.e., the expected pairwise coectivity (EPC) i the residual etworks are show i Figure. The lower EPC value, the better the algorithm performs. Thus both ad are much better tha the adhoc heuristics based o cetrality. The results of ad are highly similar, except for the largest test cases, where shows a slight advatage over. The ruig time of the algorithms are show i Figure 3 i log-scale. There is o doubt that heuristics based o cetrality takes oly a fractio of secod to complete ad is much faster tha ad. rus much slower tha, up to times slower. This expected behavior is due to the larger size of the liear program that has to cope with. Overall, turs out to be the best choice i terms of both quality ad ruig time. It rus much faster tha, ad also provide much better solutio quality tha the aive cetrality-based heuristics. VII. CONCLUSION Assessig vulerability i etworks with ucertaity is a challegig problem. While the NP-hardess of exact computatio for etwork reliability measures is a sigificat obstacle, such obstacle ca be overcome with efficiet computatioal methods, e.g., the FPRAS to compute EPC. I future, we aim to ivestigate efficiet methods to compute other etwork reliability measures as well as desig more efficiet solutios for the vulerability assesmet i forms of optimizatio problems. VIII. ACKNOWLEDGEMENT This work is partially supported by the NSF CAREER Award ad by the DTRA grat HDTRA A. Proof of Propositio APPENDIX Proof: We prove the lower ad upper bouds separately. Lower boud: By Lemma, we have EPC(G) = REL uv (G) u,v V ;u v REL uv (G) (u,v) E (u,v) E p uv Upper boud: First, we show that EPC(G) (+p e). The we ca apply the iequality of arithmetic ad geometric meas for positive umbers ( + p e ) e E to obtai EPC(G) ( ( ) m + p e ) + p e. m We prove EPC(G) ( + p e) by iductio o µ E the umber of udetermied edges (those with probabilities strictly less tha oe). Basis: If µ E =, we have a determiistic graph with m = E edges. Sice, the size of the largest compoet caot exceed m +, the pairwise coectivity is at most /(m + ) < /m(m + ) < m m. Thus, the iequality holds for µ E =. Iductio step: Assume that the iequality holds for µ E = t, we show that the iequality also holds whe µ E = t+. Assume that µ E = t+, select a arbitrary udetermied edge (u, v) E ad perform a brachig procedure o (u, v) we have EPC(G) = p uv EPC(G + ) + ( p uv )EPC(G ), where G + is obtaied from G by settig the (u, v) s probability to oe ad G is obtaied from G by removig (u, v). Sice, both G + ad G have exactly µ E udetermied edges, we ca apply the iductio hypothesis to obtai EPC(G) p uv ( + ) ( + p e ) + ( p uv ) e (u,v) e (u,v) ( + p e ) = ( + p uv ) + p e ) = e (u,v)( ( + p e ). Thus, the iequality holds for all µ E. B. Proof of Propositio Iequalities o q : O oe had q = ( p e ) p e = P E. (7) O the other had, we have q ( p e m )m (AM-GM iequality) = ( ( P E m )m/p E ) P E < exp( PE ). (8) The last step holds due to the iequality ( x) /x exp( x), x (, ),. Iequalities o q : Iterate through all edges i E, we have q = p e p e ( p e ) = q. p e e e Sice p e P E <, it follows that p e p e p e. p e P E Hece P E q P E q q. P E Iequalities o m k= q k: Apply () ad the (), we obtai m q k = q q q ( + P E ) k= ( P E )( + P E ) = P E.

9 C. Proof of Theorem 3 Proof: I step 5 of CSP, each vertex u is chose uiformly with a probability /. The BFS procedure starts from u ad gradually reveals the availability of edges whe eeded. For ay vertex v visited by the BFS procedure (icludig the startig vertex u), it takes O(d v ) times to check the availability of the icidet edges. Thus the expected umber of edges that are icidet at v ad visited by the BFS procedure at u will be REL u,v (G)d v. Ad the expected umber of visited edges by the BFS procedure at u will be v V REL u,v(g)d v. Sice u is chose uiformly, the expected umber of visited edges by the BFS procedure is REL u,v d v = ( du u V v V u V v u REL u,v ) + m From (6), the expected umber vertices visited by the BFS procedure is EPC(G)/ +. Thus the expected time complexity of the BFS procedure is O ( ) m du REL u,v + + EPC(G) u V v u Apply the iequality d u, the expected time complexity of the BFS procedure ca be simplified to ( O EPC(G) + m ). Multiply the above with N(ɛ, δ) gives us the expected time complexity of CSP O ( ɛ ( + m ) EPC(G) ). Sice the BFS procedure takes at most O(m+) time, aother upper-boud for the expected time complexity of CSP is O(N(ɛ, δ)(m + )). The combiatio ( of the above two complexity forms ) of CSP gives us the O ɛ mi{ + m EPC(G), m EPC(G) } expected time complexity of CSP. REFERENCES [] S. Neumayer, G. Zussma, R. Cohe, ad E. Modiao, Assessig the vulerability of the fiber ifrastructure to disasters, i Proc. of IEEE INFOCOM, 9. [] A. Piar, J. Meza, V. Dode, ad B. Lesieutre, Optimizatio strategies for the vulerability aalysis of the electric power grid, SIAM J. o Optimizatio, vol.,. [3] A. Murray, T. Matisziw, ad T. Grubesic, Multimethodological approaches to etwork vulerability aalysis, Growth Chage, 8. [4] A. Arulselva, C. W. Commader, L. Elefteriadou, ad P. M. Pardalos, Detectig critical odes i sparse graphs, Computers ad Operatios Research, vol. 36, o. 7, 9. [5] T. N. Dih, Y. X., M. T. Thai, E. Park, ad T. Zati, O approximatio of ew optimizatio methods for assessig etwork vulerability, i Proc. of IEEE INFOCOM,. [6] S. Neumayer, G. Zussma, R. Cohe, ad E. Modiao, Assessig the vulerability of the fiber ifrastructure to disasters, IEEE/ACM Tras. Netw., pp. 6 63,. [7] C. Colbour, The combiatorics of etwork reliability, ser. Iteratioal Series of Moographs o Computer Sciece Series. Oxford Uiversity Press, Icorporated, 987. [8] D. R. Karger, A radomized fully polyomial time approximatio scheme for the all termial etwork reliability problem, i SIAM J. COMPUT, 996, pp. 7. [9] J. F. Beders, Partitioig procedures for solvig mixed-variables programmig problems, Num. Math., vol. 4, pp. 38 5, 96. [] A. Kleywegt, A. Shapiro, ad T. Homem-de Mello, The sample average approximatio method for stochastic discrete optimizatio, SIAM Joural o Optimizatio, vol., o., pp ,. [] R. Albert, H. Jeog, ad A. Barabasi, Error ad attack tolerace of complex etworks, Nature, vol. 46, o. 6794, p. 4,. [] T. H. Grubesic, T. C. Matisziw, A. T. Murray, ad D. Sediker, Comparative approaches for assessig etwork vulerability, Iter. Regioal Sci. Review, 8. [3] T. C. Matisziw ad A. T. Murray, Modelig s-t path availability to support disaster vulerability assessmet of etwork ifrastructure, Comput. Oper. Res., vol. 36, pp. 6 6, Jauary 9. [4] T. N. Dih ad M. T. Thai, Precise structural vulerability assessmet via mathematical programmig, i Proc. of IEEE MILCOM,. [5] T. Dih ad M. Thai, Network uder joit ode ad lik attacks: Vulerability assessmet methods ad aalysis, Networkig, IEEE/ACM Trasactios o, 5. [6] K. Aggarwal, J. S. Gupta, ad K. Misra, A simple method for reliability evaluatio of a commuicatio system, Commuicatios, IEEE Trasactios o, vol. 3, o. 5, pp , 975. [7] N. S. Fard ad T.-H. Lee, Cutset eumeratio of etwork systems with lik ad ode failures, Reliability Egieerig & System Safety, vol. 65, o., pp. 4 46, 999. [8] J. Prova ad M. Ball, The complexity of coutig cuts ad of computig the probability that a graph is coected, SIAM Joural o Computig, vol., o. 4, pp , 983. [9] C. J. Colbour, Aalysis ad sythesis problems for etwork resiliece, Mathematical ad Computer Modellig, vol. 7, o., pp , 993. [] A. T. Ami, K. T. Siegrist, ad P. J. Slater, O the oexistece of uiformly optimal graphs for pair-coected reliability, Networks, vol., o. 3, pp , 99. [] S. Neumayer ad E. Modiao, Network reliability with geographically correlated failures, i INFOCOM, Proceedigs IEEE, March, pp. 9. [] P. K. Agarwal, A. Efrat, S. K. Gajugute, D. Hay, S. Sakararama, ad G. Zussma, The resiliece of wdm etworks to probabilistic geographical failures, Networkig, IEEE/ACM Trasactios o, vol., o. 5, pp , 3. [3] P. Erdos ad A. Reyi, O the evolutio of radom graphs, Publ. Math. Ist. Hugary. Acad. Sci., vol. 5, pp. 7 6, 96. [4] A. Arulselva, C. W. Commader, L. Elefteriadou, ad P. M. Pardalos, Detectig critical odes i sparse graphs, Computers & Operatios Research, vol. 36, o. 7, pp. 93, 9. [5] A. Shapiro, D. Detcheva, ad A. Ruszczyński, Lectures o Stochastic Programmig: Modelig ad Theory, ser. MPS-SIAM Series o Optimizatio Series, 9. [6] Y. She, N. P. Nguye, Y. Xua, ad M. T. Thai, O the discovery of critical liks ad odes for assessig etwork vulerability, IEEE/ACM Trasactios o Networkig (TON), vol., o. 3, pp , 3. [7] P. Dagum, R. Karp, M. Luby, ad S. Ross, A optimal algorithm for mote carlo estimatio, SIAM Joural o Computig, vol. 9, o. 5, pp ,. [8] A. Barabasi, R. Albert, ad H. Jeog, Scale-free characteristics of radom etworks: the topology of the world-wide web, Phy. A,. [9] D. J. Watts ad S. H. Strogatz, Collective dyamics of small-world etworks. Nature, vol. 393, o. 6684, 998. [3] US IP Backboe etwork XO compay,

Modified Line Search Method for Global Optimization

Modified Line Search Method for Global Optimization Modified Lie Search Method for Global Optimizatio Cria Grosa ad Ajith Abraham Ceter of Excellece for Quatifiable Quality of Service Norwegia Uiversity of Sciece ad Techology Trodheim, Norway {cria, ajith}@q2s.tu.o

More information

Asymptotic Growth of Functions

Asymptotic Growth of Functions CMPS Itroductio to Aalysis of Algorithms Fall 3 Asymptotic Growth of Fuctios We itroduce several types of asymptotic otatio which are used to compare the performace ad efficiecy of algorithms As we ll

More information

5 Boolean Decision Trees (February 11)

5 Boolean Decision Trees (February 11) 5 Boolea Decisio Trees (February 11) 5.1 Graph Coectivity Suppose we are give a udirected graph G, represeted as a boolea adjacecy matrix = (a ij ), where a ij = 1 if ad oly if vertices i ad j are coected

More information

In nite Sequences. Dr. Philippe B. Laval Kennesaw State University. October 9, 2008

In nite Sequences. Dr. Philippe B. Laval Kennesaw State University. October 9, 2008 I ite Sequeces Dr. Philippe B. Laval Keesaw State Uiversity October 9, 2008 Abstract This had out is a itroductio to i ite sequeces. mai de itios ad presets some elemetary results. It gives the I ite Sequeces

More information

Running Time ( 3.1) Analysis of Algorithms. Experimental Studies ( 3.1.1) Limitations of Experiments. Pseudocode ( 3.1.2) Theoretical Analysis

Running Time ( 3.1) Analysis of Algorithms. Experimental Studies ( 3.1.1) Limitations of Experiments. Pseudocode ( 3.1.2) Theoretical Analysis Ruig Time ( 3.) Aalysis of Algorithms Iput Algorithm Output A algorithm is a step-by-step procedure for solvig a problem i a fiite amout of time. Most algorithms trasform iput objects ito output objects.

More information

Network under Joint Node and Link Attacks: Vulnerability Assessment Methods and Analysis

Network under Joint Node and Link Attacks: Vulnerability Assessment Methods and Analysis Network uder Joit Node ad Lik Attacks: Vulerability Assessmet Methods ad Aalysis Thag N. Dih, Member, IEEE ad My T. Thai, Member, IEEE, Abstract Critical ifrastructures such as commuicatio etworks, electrical

More information

Lecture 2: Karger s Min Cut Algorithm

Lecture 2: Karger s Min Cut Algorithm priceto uiv. F 3 cos 5: Advaced Algorithm Desig Lecture : Karger s Mi Cut Algorithm Lecturer: Sajeev Arora Scribe:Sajeev Today s topic is simple but gorgeous: Karger s mi cut algorithm ad its extesio.

More information

Chapter 6: Variance, the law of large numbers and the Monte-Carlo method

Chapter 6: Variance, the law of large numbers and the Monte-Carlo method Chapter 6: Variace, the law of large umbers ad the Mote-Carlo method Expected value, variace, ad Chebyshev iequality. If X is a radom variable recall that the expected value of X, E[X] is the average value

More information

Department of Computer Science, University of Otago

Department of Computer Science, University of Otago Departmet of Computer Sciece, Uiversity of Otago Techical Report OUCS-2006-09 Permutatios Cotaiig May Patters Authors: M.H. Albert Departmet of Computer Sciece, Uiversity of Otago Micah Colema, Rya Fly

More information

Lecture 13. Lecturer: Jonathan Kelner Scribe: Jonathan Pines (2009)

Lecture 13. Lecturer: Jonathan Kelner Scribe: Jonathan Pines (2009) 18.409 A Algorithmist s Toolkit October 27, 2009 Lecture 13 Lecturer: Joatha Keler Scribe: Joatha Pies (2009) 1 Outlie Last time, we proved the Bru-Mikowski iequality for boxes. Today we ll go over the

More information

Soving Recurrence Relations

Soving Recurrence Relations Sovig Recurrece Relatios Part 1. Homogeeous liear 2d degree relatios with costat coefficiets. Cosider the recurrece relatio ( ) T () + at ( 1) + bt ( 2) = 0 This is called a homogeeous liear 2d degree

More information

I. Chi-squared Distributions

I. Chi-squared Distributions 1 M 358K Supplemet to Chapter 23: CHI-SQUARED DISTRIBUTIONS, T-DISTRIBUTIONS, AND DEGREES OF FREEDOM To uderstad t-distributios, we first eed to look at aother family of distributios, the chi-squared distributios.

More information

Taking DCOP to the Real World: Efficient Complete Solutions for Distributed Multi-Event Scheduling

Taking DCOP to the Real World: Efficient Complete Solutions for Distributed Multi-Event Scheduling Taig DCOP to the Real World: Efficiet Complete Solutios for Distributed Multi-Evet Schedulig Rajiv T. Maheswara, Milid Tambe, Emma Bowrig, Joatha P. Pearce, ad Pradeep araatham Uiversity of Souther Califoria

More information

Discrete Mathematics and Probability Theory Spring 2014 Anant Sahai Note 13

Discrete Mathematics and Probability Theory Spring 2014 Anant Sahai Note 13 EECS 70 Discrete Mathematics ad Probability Theory Sprig 2014 Aat Sahai Note 13 Itroductio At this poit, we have see eough examples that it is worth just takig stock of our model of probability ad may

More information

Output Analysis (2, Chapters 10 &11 Law)

Output Analysis (2, Chapters 10 &11 Law) B. Maddah ENMG 6 Simulatio 05/0/07 Output Aalysis (, Chapters 10 &11 Law) Comparig alterative system cofiguratio Sice the output of a simulatio is radom, the comparig differet systems via simulatio should

More information

Systems Design Project: Indoor Location of Wireless Devices

Systems Design Project: Indoor Location of Wireless Devices Systems Desig Project: Idoor Locatio of Wireless Devices Prepared By: Bria Murphy Seior Systems Sciece ad Egieerig Washigto Uiversity i St. Louis Phoe: (805) 698-5295 Email: bcm1@cec.wustl.edu Supervised

More information

Chapter 7 Methods of Finding Estimators

Chapter 7 Methods of Finding Estimators Chapter 7 for BST 695: Special Topics i Statistical Theory. Kui Zhag, 011 Chapter 7 Methods of Fidig Estimators Sectio 7.1 Itroductio Defiitio 7.1.1 A poit estimator is ay fuctio W( X) W( X1, X,, X ) of

More information

A probabilistic proof of a binomial identity

A probabilistic proof of a binomial identity A probabilistic proof of a biomial idetity Joatho Peterso Abstract We give a elemetary probabilistic proof of a biomial idetity. The proof is obtaied by computig the probability of a certai evet i two

More information

CHAPTER 3 THE TIME VALUE OF MONEY

CHAPTER 3 THE TIME VALUE OF MONEY CHAPTER 3 THE TIME VALUE OF MONEY OVERVIEW A dollar i the had today is worth more tha a dollar to be received i the future because, if you had it ow, you could ivest that dollar ad ear iterest. Of all

More information

Project Deliverables. CS 361, Lecture 28. Outline. Project Deliverables. Administrative. Project Comments

Project Deliverables. CS 361, Lecture 28. Outline. Project Deliverables. Administrative. Project Comments Project Deliverables CS 361, Lecture 28 Jared Saia Uiversity of New Mexico Each Group should tur i oe group project cosistig of: About 6-12 pages of text (ca be loger with appedix) 6-12 figures (please

More information

INVESTMENT PERFORMANCE COUNCIL (IPC)

INVESTMENT PERFORMANCE COUNCIL (IPC) INVESTMENT PEFOMANCE COUNCIL (IPC) INVITATION TO COMMENT: Global Ivestmet Performace Stadards (GIPS ) Guidace Statemet o Calculatio Methodology The Associatio for Ivestmet Maagemet ad esearch (AIM) seeks

More information

(VCP-310) 1-800-418-6789

(VCP-310) 1-800-418-6789 Maual VMware Lesso 1: Uderstadig the VMware Product Lie I this lesso, you will first lear what virtualizatio is. Next, you ll explore the products offered by VMware that provide virtualizatio services.

More information

SECTION 1.5 : SUMMATION NOTATION + WORK WITH SEQUENCES

SECTION 1.5 : SUMMATION NOTATION + WORK WITH SEQUENCES SECTION 1.5 : SUMMATION NOTATION + WORK WITH SEQUENCES Read Sectio 1.5 (pages 5 9) Overview I Sectio 1.5 we lear to work with summatio otatio ad formulas. We will also itroduce a brief overview of sequeces,

More information

Domain 1: Designing a SQL Server Instance and a Database Solution

Domain 1: Designing a SQL Server Instance and a Database Solution Maual SQL Server 2008 Desig, Optimize ad Maitai (70-450) 1-800-418-6789 Domai 1: Desigig a SQL Server Istace ad a Database Solutio Desigig for CPU, Memory ad Storage Capacity Requiremets Whe desigig a

More information

THE ARITHMETIC OF INTEGERS. - multiplication, exponentiation, division, addition, and subtraction

THE ARITHMETIC OF INTEGERS. - multiplication, exponentiation, division, addition, and subtraction THE ARITHMETIC OF INTEGERS - multiplicatio, expoetiatio, divisio, additio, ad subtractio What to do ad what ot to do. THE INTEGERS Recall that a iteger is oe of the whole umbers, which may be either positive,

More information

Vladimir N. Burkov, Dmitri A. Novikov MODELS AND METHODS OF MULTIPROJECTS MANAGEMENT

Vladimir N. Burkov, Dmitri A. Novikov MODELS AND METHODS OF MULTIPROJECTS MANAGEMENT Keywords: project maagemet, resource allocatio, etwork plaig Vladimir N Burkov, Dmitri A Novikov MODELS AND METHODS OF MULTIPROJECTS MANAGEMENT The paper deals with the problems of resource allocatio betwee

More information

Hypergeometric Distributions

Hypergeometric Distributions 7.4 Hypergeometric Distributios Whe choosig the startig lie-up for a game, a coach obviously has to choose a differet player for each positio. Similarly, whe a uio elects delegates for a covetio or you

More information

The Power of Free Branching in a General Model of Backtracking and Dynamic Programming Algorithms

The Power of Free Branching in a General Model of Backtracking and Dynamic Programming Algorithms The Power of Free Brachig i a Geeral Model of Backtrackig ad Dyamic Programmig Algorithms SASHKA DAVIS IDA/Ceter for Computig Scieces Bowie, MD sashka.davis@gmail.com RUSSELL IMPAGLIAZZO Dept. of Computer

More information

PROCEEDINGS OF THE YEREVAN STATE UNIVERSITY AN ALTERNATIVE MODEL FOR BONUS-MALUS SYSTEM

PROCEEDINGS OF THE YEREVAN STATE UNIVERSITY AN ALTERNATIVE MODEL FOR BONUS-MALUS SYSTEM PROCEEDINGS OF THE YEREVAN STATE UNIVERSITY Physical ad Mathematical Scieces 2015, 1, p. 15 19 M a t h e m a t i c s AN ALTERNATIVE MODEL FOR BONUS-MALUS SYSTEM A. G. GULYAN Chair of Actuarial Mathematics

More information

Chair for Network Architectures and Services Institute of Informatics TU München Prof. Carle. Network Security. Chapter 2 Basics

Chair for Network Architectures and Services Institute of Informatics TU München Prof. Carle. Network Security. Chapter 2 Basics Chair for Network Architectures ad Services Istitute of Iformatics TU Müche Prof. Carle Network Security Chapter 2 Basics 2.4 Radom Number Geeratio for Cryptographic Protocols Motivatio It is crucial to

More information

Incremental calculation of weighted mean and variance

Incremental calculation of weighted mean and variance Icremetal calculatio of weighted mea ad variace Toy Fich faf@cam.ac.uk dot@dotat.at Uiversity of Cambridge Computig Service February 009 Abstract I these otes I eplai how to derive formulae for umerically

More information

MARTINGALES AND A BASIC APPLICATION

MARTINGALES AND A BASIC APPLICATION MARTINGALES AND A BASIC APPLICATION TURNER SMITH Abstract. This paper will develop the measure-theoretic approach to probability i order to preset the defiitio of martigales. From there we will apply this

More information

Chatpun Khamyat Department of Industrial Engineering, Kasetsart University, Bangkok, Thailand ocpky@hotmail.com

Chatpun Khamyat Department of Industrial Engineering, Kasetsart University, Bangkok, Thailand ocpky@hotmail.com SOLVING THE OIL DELIVERY TRUCKS ROUTING PROBLEM WITH MODIFY MULTI-TRAVELING SALESMAN PROBLEM APPROACH CASE STUDY: THE SME'S OIL LOGISTIC COMPANY IN BANGKOK THAILAND Chatpu Khamyat Departmet of Idustrial

More information

Sequences and Series

Sequences and Series CHAPTER 9 Sequeces ad Series 9.. Covergece: Defiitio ad Examples Sequeces The purpose of this chapter is to itroduce a particular way of geeratig algorithms for fidig the values of fuctios defied by their

More information

A Faster Clause-Shortening Algorithm for SAT with No Restriction on Clause Length

A Faster Clause-Shortening Algorithm for SAT with No Restriction on Clause Length Joural o Satisfiability, Boolea Modelig ad Computatio 1 2005) 49-60 A Faster Clause-Shorteig Algorithm for SAT with No Restrictio o Clause Legth Evgey Datsi Alexader Wolpert Departmet of Computer Sciece

More information

Recovery time guaranteed heuristic routing for improving computation complexity in survivable WDM networks

Recovery time guaranteed heuristic routing for improving computation complexity in survivable WDM networks Computer Commuicatios 30 (2007) 1331 1336 wwwelseviercom/locate/comcom Recovery time guarateed heuristic routig for improvig computatio complexity i survivable WDM etworks Lei Guo * College of Iformatio

More information

Irreducible polynomials with consecutive zero coefficients

Irreducible polynomials with consecutive zero coefficients Irreducible polyomials with cosecutive zero coefficiets Theodoulos Garefalakis Departmet of Mathematics, Uiversity of Crete, 71409 Heraklio, Greece Abstract Let q be a prime power. We cosider the problem

More information

where: T = number of years of cash flow in investment's life n = the year in which the cash flow X n i = IRR = the internal rate of return

where: T = number of years of cash flow in investment's life n = the year in which the cash flow X n i = IRR = the internal rate of return EVALUATING ALTERNATIVE CAPITAL INVESTMENT PROGRAMS By Ke D. Duft, Extesio Ecoomist I the March 98 issue of this publicatio we reviewed the procedure by which a capital ivestmet project was assessed. The

More information

1 Computing the Standard Deviation of Sample Means

1 Computing the Standard Deviation of Sample Means Computig the Stadard Deviatio of Sample Meas Quality cotrol charts are based o sample meas ot o idividual values withi a sample. A sample is a group of items, which are cosidered all together for our aalysis.

More information

Properties of MLE: consistency, asymptotic normality. Fisher information.

Properties of MLE: consistency, asymptotic normality. Fisher information. Lecture 3 Properties of MLE: cosistecy, asymptotic ormality. Fisher iformatio. I this sectio we will try to uderstad why MLEs are good. Let us recall two facts from probability that we be used ofte throughout

More information

LECTURE 13: Cross-validation

LECTURE 13: Cross-validation LECTURE 3: Cross-validatio Resampli methods Cross Validatio Bootstrap Bias ad variace estimatio with the Bootstrap Three-way data partitioi Itroductio to Patter Aalysis Ricardo Gutierrez-Osua Texas A&M

More information

Your organization has a Class B IP address of 166.144.0.0 Before you implement subnetting, the Network ID and Host ID are divided as follows:

Your organization has a Class B IP address of 166.144.0.0 Before you implement subnetting, the Network ID and Host ID are divided as follows: Subettig Subettig is used to subdivide a sigle class of etwork i to multiple smaller etworks. Example: Your orgaizatio has a Class B IP address of 166.144.0.0 Before you implemet subettig, the Network

More information

Estimating Probability Distributions by Observing Betting Practices

Estimating Probability Distributions by Observing Betting Practices 5th Iteratioal Symposium o Imprecise Probability: Theories ad Applicatios, Prague, Czech Republic, 007 Estimatig Probability Distributios by Observig Bettig Practices Dr C Lych Natioal Uiversity of Irelad,

More information

Lecture 4: Cheeger s Inequality

Lecture 4: Cheeger s Inequality Spectral Graph Theory ad Applicatios WS 0/0 Lecture 4: Cheeger s Iequality Lecturer: Thomas Sauerwald & He Su Statemet of Cheeger s Iequality I this lecture we assume for simplicity that G is a d-regular

More information

Week 3 Conditional probabilities, Bayes formula, WEEK 3 page 1 Expected value of a random variable

Week 3 Conditional probabilities, Bayes formula, WEEK 3 page 1 Expected value of a random variable Week 3 Coditioal probabilities, Bayes formula, WEEK 3 page 1 Expected value of a radom variable We recall our discussio of 5 card poker hads. Example 13 : a) What is the probability of evet A that a 5

More information

Measures of Spread and Boxplots Discrete Math, Section 9.4

Measures of Spread and Boxplots Discrete Math, Section 9.4 Measures of Spread ad Boxplots Discrete Math, Sectio 9.4 We start with a example: Example 1: Comparig Mea ad Media Compute the mea ad media of each data set: S 1 = {4, 6, 8, 10, 1, 14, 16} S = {4, 7, 9,

More information

THE ABRACADABRA PROBLEM

THE ABRACADABRA PROBLEM THE ABRACADABRA PROBLEM FRANCESCO CARAVENNA Abstract. We preset a detailed solutio of Exercise E0.6 i [Wil9]: i a radom sequece of letters, draw idepedetly ad uiformly from the Eglish alphabet, the expected

More information

INVESTMENT PERFORMANCE COUNCIL (IPC) Guidance Statement on Calculation Methodology

INVESTMENT PERFORMANCE COUNCIL (IPC) Guidance Statement on Calculation Methodology Adoptio Date: 4 March 2004 Effective Date: 1 Jue 2004 Retroactive Applicatio: No Public Commet Period: Aug Nov 2002 INVESTMENT PERFORMANCE COUNCIL (IPC) Preface Guidace Statemet o Calculatio Methodology

More information

On the Capacity of Hybrid Wireless Networks

On the Capacity of Hybrid Wireless Networks O the Capacity of Hybrid ireless Networks Beyua Liu,ZheLiu +,DoTowsley Departmet of Computer Sciece Uiversity of Massachusetts Amherst, MA 0002 + IBM T.J. atso Research Ceter P.O. Box 704 Yorktow Heights,

More information

3 Basic Definitions of Probability Theory

3 Basic Definitions of Probability Theory 3 Basic Defiitios of Probability Theory 3defprob.tex: Feb 10, 2003 Classical probability Frequecy probability axiomatic probability Historical developemet: Classical Frequecy Axiomatic The Axiomatic defiitio

More information

TIGHT BOUNDS ON EXPECTED ORDER STATISTICS

TIGHT BOUNDS ON EXPECTED ORDER STATISTICS Probability i the Egieerig ad Iformatioal Scieces, 20, 2006, 667 686+ Prited i the U+S+A+ TIGHT BOUNDS ON EXPECTED ORDER STATISTICS DIMITRIS BERTSIMAS Sloa School of Maagemet ad Operatios Research Ceter

More information

*The most important feature of MRP as compared with ordinary inventory control analysis is its time phasing feature.

*The most important feature of MRP as compared with ordinary inventory control analysis is its time phasing feature. Itegrated Productio ad Ivetory Cotrol System MRP ad MRP II Framework of Maufacturig System Ivetory cotrol, productio schedulig, capacity plaig ad fiacial ad busiess decisios i a productio system are iterrelated.

More information

Hypothesis testing. Null and alternative hypotheses

Hypothesis testing. Null and alternative hypotheses Hypothesis testig Aother importat use of samplig distributios is to test hypotheses about populatio parameters, e.g. mea, proportio, regressio coefficiets, etc. For example, it is possible to stipulate

More information

Present Values, Investment Returns and Discount Rates

Present Values, Investment Returns and Discount Rates Preset Values, Ivestmet Returs ad Discout Rates Dimitry Midli, ASA, MAAA, PhD Presidet CDI Advisors LLC dmidli@cdiadvisors.com May 2, 203 Copyright 20, CDI Advisors LLC The cocept of preset value lies

More information

Center, Spread, and Shape in Inference: Claims, Caveats, and Insights

Center, Spread, and Shape in Inference: Claims, Caveats, and Insights Ceter, Spread, ad Shape i Iferece: Claims, Caveats, ad Isights Dr. Nacy Pfeig (Uiversity of Pittsburgh) AMATYC November 2008 Prelimiary Activities 1. I would like to produce a iterval estimate for the

More information

NEW HIGH PERFORMANCE COMPUTATIONAL METHODS FOR MORTGAGES AND ANNUITIES. Yuri Shestopaloff,

NEW HIGH PERFORMANCE COMPUTATIONAL METHODS FOR MORTGAGES AND ANNUITIES. Yuri Shestopaloff, NEW HIGH PERFORMNCE COMPUTTIONL METHODS FOR MORTGGES ND NNUITIES Yuri Shestopaloff, Geerally, mortgage ad auity equatios do ot have aalytical solutios for ukow iterest rate, which has to be foud usig umerical

More information

Chapter 5 O A Cojecture Of Erdíos Proceedigs NCUR VIII è1994è, Vol II, pp 794í798 Jeærey F Gold Departmet of Mathematics, Departmet of Physics Uiversity of Utah Do H Tucker Departmet of Mathematics Uiversity

More information

5: Introduction to Estimation

5: Introduction to Estimation 5: Itroductio to Estimatio Cotets Acroyms ad symbols... 1 Statistical iferece... Estimatig µ with cofidece... 3 Samplig distributio of the mea... 3 Cofidece Iterval for μ whe σ is kow before had... 4 Sample

More information

CS103A Handout 23 Winter 2002 February 22, 2002 Solving Recurrence Relations

CS103A Handout 23 Winter 2002 February 22, 2002 Solving Recurrence Relations CS3A Hadout 3 Witer 00 February, 00 Solvig Recurrece Relatios Itroductio A wide variety of recurrece problems occur i models. Some of these recurrece relatios ca be solved usig iteratio or some other ad

More information

Class Meeting # 16: The Fourier Transform on R n

Class Meeting # 16: The Fourier Transform on R n MATH 18.152 COUSE NOTES - CLASS MEETING # 16 18.152 Itroductio to PDEs, Fall 2011 Professor: Jared Speck Class Meetig # 16: The Fourier Trasform o 1. Itroductio to the Fourier Trasform Earlier i the course,

More information

THE HEIGHT OF q-binary SEARCH TREES

THE HEIGHT OF q-binary SEARCH TREES THE HEIGHT OF q-binary SEARCH TREES MICHAEL DRMOTA AND HELMUT PRODINGER Abstract. q biary search trees are obtaied from words, equipped with the geometric distributio istead of permutatios. The average

More information

A Combined Continuous/Binary Genetic Algorithm for Microstrip Antenna Design

A Combined Continuous/Binary Genetic Algorithm for Microstrip Antenna Design A Combied Cotiuous/Biary Geetic Algorithm for Microstrip Atea Desig Rady L. Haupt The Pesylvaia State Uiversity Applied Research Laboratory P. O. Box 30 State College, PA 16804-0030 haupt@ieee.org Abstract:

More information

Tradigms of Astundithi and Toyota

Tradigms of Astundithi and Toyota Tradig the radomess - Desigig a optimal tradig strategy uder a drifted radom walk price model Yuao Wu Math 20 Project Paper Professor Zachary Hamaker Abstract: I this paper the author iteds to explore

More information

.04. This means $1000 is multiplied by 1.02 five times, once for each of the remaining sixmonth

.04. This means $1000 is multiplied by 1.02 five times, once for each of the remaining sixmonth Questio 1: What is a ordiary auity? Let s look at a ordiary auity that is certai ad simple. By this, we mea a auity over a fixed term whose paymet period matches the iterest coversio period. Additioally,

More information

Non-life insurance mathematics. Nils F. Haavardsson, University of Oslo and DNB Skadeforsikring

Non-life insurance mathematics. Nils F. Haavardsson, University of Oslo and DNB Skadeforsikring No-life isurace mathematics Nils F. Haavardsso, Uiversity of Oslo ad DNB Skadeforsikrig Mai issues so far Why does isurace work? How is risk premium defied ad why is it importat? How ca claim frequecy

More information

Here are a couple of warnings to my students who may be here to get a copy of what happened on a day that you missed.

Here are a couple of warnings to my students who may be here to get a copy of what happened on a day that you missed. This documet was writte ad copyrighted by Paul Dawkis. Use of this documet ad its olie versio is govered by the Terms ad Coditios of Use located at http://tutorial.math.lamar.edu/terms.asp. The olie versio

More information

Analyzing Longitudinal Data from Complex Surveys Using SUDAAN

Analyzing Longitudinal Data from Complex Surveys Using SUDAAN Aalyzig Logitudial Data from Complex Surveys Usig SUDAAN Darryl Creel Statistics ad Epidemiology, RTI Iteratioal, 312 Trotter Farm Drive, Rockville, MD, 20850 Abstract SUDAAN: Software for the Statistical

More information

A Constant-Factor Approximation Algorithm for the Link Building Problem

A Constant-Factor Approximation Algorithm for the Link Building Problem A Costat-Factor Approximatio Algorithm for the Lik Buildig Problem Marti Olse 1, Aastasios Viglas 2, ad Ilia Zvedeiouk 2 1 Ceter for Iovatio ad Busiess Developmet, Istitute of Busiess ad Techology, Aarhus

More information

CS100: Introduction to Computer Science

CS100: Introduction to Computer Science Review: History of Computers CS100: Itroductio to Computer Sciece Maiframes Miicomputers Lecture 2: Data Storage -- Bits, their storage ad mai memory Persoal Computers & Workstatios Review: The Role of

More information

Accurate and Efficient Traffic Monitoring Using Adaptive Non-linear Sampling Method

Accurate and Efficient Traffic Monitoring Using Adaptive Non-linear Sampling Method Accurate ad Efficiet Traffic Moitorig Usig Adaptive No-liear Samplig Method Chegche Hu, Sheg Wag, Jia Tia, Bi Liu Tsighua Uiversity Beijig, Chia, {hucc,wags,tiaj}@mails.tsighua.edu.c liub@tsighua.edu.c

More information

Universal coding for classes of sources

Universal coding for classes of sources Coexios module: m46228 Uiversal codig for classes of sources Dever Greee This work is produced by The Coexios Project ad licesed uder the Creative Commos Attributio Licese We have discussed several parametric

More information

Chapter 7 - Sampling Distributions. 1 Introduction. What is statistics? It consist of three major areas:

Chapter 7 - Sampling Distributions. 1 Introduction. What is statistics? It consist of three major areas: Chapter 7 - Samplig Distributios 1 Itroductio What is statistics? It cosist of three major areas: Data Collectio: samplig plas ad experimetal desigs Descriptive Statistics: umerical ad graphical summaries

More information

Lecture 3. denote the orthogonal complement of S k. Then. 1 x S k. n. 2 x T Ax = ( ) λ x. with x = 1, we have. i = λ k x 2 = λ k.

Lecture 3. denote the orthogonal complement of S k. Then. 1 x S k. n. 2 x T Ax = ( ) λ x. with x = 1, we have. i = λ k x 2 = λ k. 18.409 A Algorithmist s Toolkit September 17, 009 Lecture 3 Lecturer: Joatha Keler Scribe: Adre Wibisoo 1 Outlie Today s lecture covers three mai parts: Courat-Fischer formula ad Rayleigh quotiets The

More information

ADAPTIVE NETWORKS SAFETY CONTROL ON FUZZY LOGIC

ADAPTIVE NETWORKS SAFETY CONTROL ON FUZZY LOGIC 8 th Iteratioal Coferece o DEVELOPMENT AND APPLICATION SYSTEMS S u c e a v a, R o m a i a, M a y 25 27, 2 6 ADAPTIVE NETWORKS SAFETY CONTROL ON FUZZY LOGIC Vadim MUKHIN 1, Elea PAVLENKO 2 Natioal Techical

More information

Institute of Actuaries of India Subject CT1 Financial Mathematics

Institute of Actuaries of India Subject CT1 Financial Mathematics Istitute of Actuaries of Idia Subject CT1 Fiacial Mathematics For 2014 Examiatios Subject CT1 Fiacial Mathematics Core Techical Aim The aim of the Fiacial Mathematics subject is to provide a groudig i

More information

Heterogeneous Vehicle Routing Problem with profits Dynamic solving by Clustering Genetic Algorithm

Heterogeneous Vehicle Routing Problem with profits Dynamic solving by Clustering Genetic Algorithm IJCSI Iteratioal Joural of Computer Sciece Issues, Vol. 10, Issue 4, No 1, July 2013 ISSN (Prit): 1694-0814 ISSN (Olie): 1694-0784 www.ijcsi.org 247 Heterogeeous Vehicle Routig Problem with profits Dyamic

More information

Convexity, Inequalities, and Norms

Convexity, Inequalities, and Norms Covexity, Iequalities, ad Norms Covex Fuctios You are probably familiar with the otio of cocavity of fuctios. Give a twicedifferetiable fuctio ϕ: R R, We say that ϕ is covex (or cocave up) if ϕ (x) 0 for

More information

Perfect Packing Theorems and the Average-Case Behavior of Optimal and Online Bin Packing

Perfect Packing Theorems and the Average-Case Behavior of Optimal and Online Bin Packing SIAM REVIEW Vol. 44, No. 1, pp. 95 108 c 2002 Society for Idustrial ad Applied Mathematics Perfect Packig Theorems ad the Average-Case Behavior of Optimal ad Olie Bi Packig E. G. Coffma, Jr. C. Courcoubetis

More information

Subject CT5 Contingencies Core Technical Syllabus

Subject CT5 Contingencies Core Technical Syllabus Subject CT5 Cotigecies Core Techical Syllabus for the 2015 exams 1 Jue 2014 Aim The aim of the Cotigecies subject is to provide a groudig i the mathematical techiques which ca be used to model ad value

More information

Approximating Area under a curve with rectangles. To find the area under a curve we approximate the area using rectangles and then use limits to find

Approximating Area under a curve with rectangles. To find the area under a curve we approximate the area using rectangles and then use limits to find 1.8 Approximatig Area uder a curve with rectagles 1.6 To fid the area uder a curve we approximate the area usig rectagles ad the use limits to fid 1.4 the area. Example 1 Suppose we wat to estimate 1.

More information

Entropy of bi-capacities

Entropy of bi-capacities Etropy of bi-capacities Iva Kojadiovic LINA CNRS FRE 2729 Site école polytechique de l uiv. de Nates Rue Christia Pauc 44306 Nates, Frace iva.kojadiovic@uiv-ates.fr Jea-Luc Marichal Applied Mathematics

More information

Multiple Representations for Pattern Exploration with the Graphing Calculator and Manipulatives

Multiple Representations for Pattern Exploration with the Graphing Calculator and Manipulatives Douglas A. Lapp Multiple Represetatios for Patter Exploratio with the Graphig Calculator ad Maipulatives To teach mathematics as a coected system of cocepts, we must have a shift i emphasis from a curriculum

More information

Stochastic Online Scheduling with Precedence Constraints

Stochastic Online Scheduling with Precedence Constraints Stochastic Olie Schedulig with Precedece Costraits Nicole Megow Tark Vredeveld July 15, 2008 Abstract We cosider the preemptive ad o-preemptive problems of schedulig obs with precedece costraits o parallel

More information

Design of Auctions for Electronic Business

Design of Auctions for Electronic Business Desig of Auctios for Electroic Busiess Article Ifo: Maagemet Iformatio Systems, Vol. 5 (200), No., pp. 037-042 Received 8 September 2009 Accepted 7 April 200 UDC 004.738.5:339]::005; 005.52/.53 ; 347.45.6

More information

MTO-MTS Production Systems in Supply Chains

MTO-MTS Production Systems in Supply Chains NSF GRANT #0092854 NSF PROGRAM NAME: MES/OR MTO-MTS Productio Systems i Supply Chais Philip M. Kamisky Uiversity of Califoria, Berkeley Our Kaya Uiversity of Califoria, Berkeley Abstract: Icreasig cost

More information

Case Study. Normal and t Distributions. Density Plot. Normal Distributions

Case Study. Normal and t Distributions. Density Plot. Normal Distributions Case Study Normal ad t Distributios Bret Halo ad Bret Larget Departmet of Statistics Uiversity of Wiscosi Madiso October 11 13, 2011 Case Study Body temperature varies withi idividuals over time (it ca

More information

Example 2 Find the square root of 0. The only square root of 0 is 0 (since 0 is not positive or negative, so those choices don t exist here).

Example 2 Find the square root of 0. The only square root of 0 is 0 (since 0 is not positive or negative, so those choices don t exist here). BEGINNING ALGEBRA Roots ad Radicals (revised summer, 00 Olso) Packet to Supplemet the Curret Textbook - Part Review of Square Roots & Irratioals (This portio ca be ay time before Part ad should mostly

More information

Plug-in martingales for testing exchangeability on-line

Plug-in martingales for testing exchangeability on-line Plug-i martigales for testig exchageability o-lie Valetia Fedorova, Alex Gammerma, Ilia Nouretdiov, ad Vladimir Vovk Computer Learig Research Cetre Royal Holloway, Uiversity of Lodo, UK {valetia,ilia,alex,vovk}@cs.rhul.ac.uk

More information

Normal Distribution.

Normal Distribution. Normal Distributio www.icrf.l Normal distributio I probability theory, the ormal or Gaussia distributio, is a cotiuous probability distributio that is ofte used as a first approimatio to describe realvalued

More information

Foundations of Operations Research

Foundations of Operations Research Foudatios of Operatios Research Master of Sciece i Computer Egieerig Roberto Cordoe roberto.cordoe@uimi.it Tuesday 13.15-15.15 Thursday 10.15-13.15 http://homes.di.uimi.it/~cordoe/courses/2014-for/2014-for.html

More information

COMPARISON OF THE EFFICIENCY OF S-CONTROL CHART AND EWMA-S 2 CONTROL CHART FOR THE CHANGES IN A PROCESS

COMPARISON OF THE EFFICIENCY OF S-CONTROL CHART AND EWMA-S 2 CONTROL CHART FOR THE CHANGES IN A PROCESS COMPARISON OF THE EFFICIENCY OF S-CONTROL CHART AND EWMA-S CONTROL CHART FOR THE CHANGES IN A PROCESS Supraee Lisawadi Departmet of Mathematics ad Statistics, Faculty of Sciece ad Techoology, Thammasat

More information

Infinite Sequences and Series

Infinite Sequences and Series CHAPTER 4 Ifiite Sequeces ad Series 4.1. Sequeces A sequece is a ifiite ordered list of umbers, for example the sequece of odd positive itegers: 1, 3, 5, 7, 9, 11, 13, 15, 17, 19, 21, 23, 25, 27, 29...

More information

0.7 0.6 0.2 0 0 96 96.5 97 97.5 98 98.5 99 99.5 100 100.5 96.5 97 97.5 98 98.5 99 99.5 100 100.5

0.7 0.6 0.2 0 0 96 96.5 97 97.5 98 98.5 99 99.5 100 100.5 96.5 97 97.5 98 98.5 99 99.5 100 100.5 Sectio 13 Kolmogorov-Smirov test. Suppose that we have a i.i.d. sample X 1,..., X with some ukow distributio P ad we would like to test the hypothesis that P is equal to a particular distributio P 0, i.e.

More information

Concept: Types of algorithms

Concept: Types of algorithms Discrete Math for Bioiformatics WS 10/11:, by A. Bockmayr/K. Reiert, 18. Oktober 2010, 21:22 1001 Cocept: Types of algorithms The expositio is based o the followig sources, which are all required readig:

More information

Lecture 4: Cauchy sequences, Bolzano-Weierstrass, and the Squeeze theorem

Lecture 4: Cauchy sequences, Bolzano-Weierstrass, and the Squeeze theorem Lecture 4: Cauchy sequeces, Bolzao-Weierstrass, ad the Squeeze theorem The purpose of this lecture is more modest tha the previous oes. It is to state certai coditios uder which we are guarateed that limits

More information

The Stable Marriage Problem

The Stable Marriage Problem The Stable Marriage Problem William Hut Lae Departmet of Computer Sciece ad Electrical Egieerig, West Virgiia Uiversity, Morgatow, WV William.Hut@mail.wvu.edu 1 Itroductio Imagie you are a matchmaker,

More information

Lesson 15 ANOVA (analysis of variance)

Lesson 15 ANOVA (analysis of variance) Outlie Variability -betwee group variability -withi group variability -total variability -F-ratio Computatio -sums of squares (betwee/withi/total -degrees of freedom (betwee/withi/total -mea square (betwee/withi

More information

Dynamic House Allocation

Dynamic House Allocation Dyamic House Allocatio Sujit Gujar 1 ad James Zou 2 ad David C. Parkes 3 Abstract. We study a dyamic variat o the house allocatio problem. Each aget ows a distict object (a house) ad is able to trade its

More information

Designing Incentives for Online Question and Answer Forums

Designing Incentives for Online Question and Answer Forums Desigig Icetives for Olie Questio ad Aswer Forums Shaili Jai School of Egieerig ad Applied Scieces Harvard Uiversity Cambridge, MA 0238 USA shailij@eecs.harvard.edu Yilig Che School of Egieerig ad Applied

More information

Capacity of Wireless Networks with Heterogeneous Traffic

Capacity of Wireless Networks with Heterogeneous Traffic Capacity of Wireless Networks with Heterogeeous Traffic Migyue Ji, Zheg Wag, Hamid R. Sadjadpour, J.J. Garcia-Lua-Aceves Departmet of Electrical Egieerig ad Computer Egieerig Uiversity of Califoria, Sata

More information