Simultaneous Perturbation Stochastic Approximation in Decentralized Load Balancing Problem
|
|
|
- Elmer Page
- 10 years ago
- Views:
Transcription
1 Preprins, 1s IFAC Conference on Modelling, Idenificaion and Conrol of Nonlinear Sysems June 24-26, Sain Peersburg, Russia Simulaneous Perurbaion Sochasic Approximaion in Decenralized Load Balancing Problem Naalia Amelina, Vicoria Erofeeva, Oleg Granichin, Nikolai Malkovskii Sain Peersburg Sae Universiy (Faculy of Mahemaics and Mechanics), S. Peersburg, Russia Sain Peersburg Sae Universiy (Faculy of Mahemaics and Mechanics, and Research Laboraory for Analysis and Modeling of Social Processes), Insiue of Problems in Mechanical Engineering, Russian Academy of Sciences, and ITMO Universiy, S. Peersburg, Russia oleg Absrac: In his work he load balancing problem is sudied for decenralized sochasic nework wih unknown bu bounded noise in measuremens and varying produciviies of agens. The load balancing problem is formulaed as a consensus problem in a sochasic nework. Consideraion of Laplasian poenial funcion corresponded o he nework graph allows o inroduce a new randomized local voing proocol wih consan sep-size which is based on simulaneous perurbaion sochasic approximaion algorihm. The condiions are formulaed for he approximae consensus achievemen which corresponds o achieving of a subopimal level of agens load. The new algorihm is illusraed by simulaions. Keywords: Simulaneous perurbaion sochasic approximaion, randomized algorihms, muliagen sysems, consensus problem. 1. INTRODUCTION In recen years he consensus approach has been widely used for solving differen pracical problems Olfai-Saber and Murray (2004); Olfai-Saber e al. (2007); Ren e al. (2007); Ren and Beard (2008); Cheboarev and Agaev (2009); Kar and Moura (2009); Granichin e al. (2012); Amelin e al. (2013); Lewis e al. (2014), including he load balancing problem Amelina e al. (2015). For he problem of achieving consensus a lo of heoreical resuls were obained. In Tsisiklis e al. (1986); Huang and Manon (2009); Li and Zhang (2009) he sochasic approximaion ype algorihms were used for achieving he consensus, and heir applicabiliy under some saisical uncerainies was analyzed in Amelina and Fradkov (2012); Amelina e al. (2015), where i was assumed ha measuremen noise and delays have a saisical naure wih sandard properies of zeromean and bounded covariance. Emphasize, when he undireced opology graph has a spanning ree, he load balancing problem can be reformulaed as a minimizaion problem of a Laplacian poenial associaed wih a graph (see Olfai-Saber and Murray (2004)). In his paper we sugges o use a simulaneous perurbaion sochasic approximaion (SPSA) for solving his problem. SPSA algorihm recursively generaes esimaes along a random direcions and uses The auhors acknowledge he Russian Minisry of Educaion and Science (agreemen , unique no. RFMEFI60414X0035), RFBR (projecs , , and ), and SPbSU (projec ). only wo observaions of minimized funcion a each ieraion. SPSA and similar procedures wih one (or wo) measuremens per ieraion were inroduced in Granichin (1989, 1992) Polyak and Tsybakov (1990). and Spall (1992). They are similar o random search mehods Rasrigin (1963). The general overview of SPSA ype algorihms and heir applicaions in differen fields are done in Granichin e al. (2015). Generally, a cenralized algorihm for load balancing which is based on SPSA was considered in Granichin and Amelina (2015); Granichin (2015). The paper is organized as follows. In Secion II, he problem saemen is described, and basic conceps of a graph heory ha are used hereinafer are inroduced. In Secion III, he load balancing conrol sraegy is considered. Secion IV presens a new resul abou a mean-risk opimizaion problem under linear consrains. In Secion V we inroduce he new randomized local voing proocol and Secion VI gives condiions of an asympoic mean square ε-consensus. Simulaion resuls are given in Secion VII. Secion VIII conains conclusions. 2. PROBLEM FORMULATION Le he nework sysem be composed by m agens (processors, machines, ec.) which are numbered by naurals i, i = 1,..., m, and N = {1,..., m} be a se of agens in he sysem. This sysem execues a se of asks of he same ype. Tasks feed o he sysem in differen discree ime insans = 0, 1,... hrough differen agens. Agens perform incoming asks in Copyrigh IFAC
2 June 24-26, Sain Peersburg, Russia parallel. Tasks can be redisribued among agens based on feedbacks. We assume, ha he ask can no be inerruped afer i has been assigned o he agen. In his paper we use he following noaion and erms from he marix and graph heories. A communicaion graph (N, E) is defined by a se of nodes N and a se of edges E. A dynamic nework of d agens is deermined by a se of dynamic sysems (agens) ha inerac according o he communicaion graph. We associae a weigh a i,j > 0 wih each edge (j, i) E. A graph can be represened by an adjacency marix A = [a i,j ] wih weighs a i,j > 0 if (j, i) E, and a i,j = 0 oherwise. Assume, ha a i,i = 0. We use he noaion G A for a graph which is represened by an adjacency marix A. Define a weighed in-degree of node i as a sum of i-h row of marix A: d i (A) = n j=1 ai,j, and D(A) = diag{d i (A)} as a corresponding diagonal marix. Le L(A) = D(A) A denoes he Laplacian of he graph G A. Noe, ha he sum of rows of he Laplacian equals o zero. The symbol d max (A) sands for a maximum in-degree of he graph G A, Re(λ 2 (A)) is he real par of he second eigenvalue of marix A ordered by absolue magniude, A T is he ranspose marix. Le N i = {j : a i,j > 0} be a neighbors se of agen i N, N i is a corresponding number of neighbors. The graph G A is called undireced if a i,j = a j,i for all i, j N. A each ime insan he behavior of each agen i N is described by wo characerisics: q i is he queue lengh of aomic elemenary asks of agen i a ime insan ; θ i is he produciviy of agen i a ime insan. Here and below, an upper index of agen i is used as a corresponding number of an agen (no as an exponen). The execuion ime of a ask varies from one agen o anoher and depends on a produciviy of an agen. Consider he case when he dynamic model of he sysem is described by he following equaions q i +1 = q i θ i + z i + u i, i N, = 0, 1,..., (1) where z i are amouns of new sysem asks received hrough agen i a ime insan ; u i R are conrol acions (redisribued asks o agen i a ime insan pars of sysem asks previously received hrough oher agens), which could (and should) be chosen. We assume, ha o form he conrol sraegy u i each agen i N has knowledge abou is own produciviy, produciviies of is neighbors and noisy daa abou is own queue lengh: y i,i = q i + ξ i,i, (2) and, if he neighbors se N i is no empy, he knowledge abou produciviies of is neighbors and noisy observaions abou is neighbors queue lenghs: y i,j where {w i,j } is an observaion noise. = q j + ξ i,j, j N i, (3) Denoe T i as a ime momen when agen i complees currenly assigned asks (a ime momen ). T i can be formally described as: T i = min τ τ θk i q. i k= Consider he problem of minimizaion of implemenaion ime of all asks: max T i (q0, i u i 1, z1, i u i 2, z2, i...) min. (4) i {1,...,m} u 1 1,...,um 1,u1 2,... For he saionary case when z i = 0 (i.e. here are no new receiving asks for > 0), such value does no vary over ime and so he problem becomes a wors-case opimizaion problem (moreover, i is easy o show ha he problem can be furher reduced o minimizaion of some good convex funcional). For he nonsaionary case he problem is more difficul as we should race drifing minimum poin. 3. LOAD BALANCING An ideal scheduling algorihm is he one which keeps all he nodes busy execuing essenial asks, and minimizes he inernode communicaion required o deermine he schedule and pass daa beween asks. The scheduling problem is paricularly challenging when he asks are generaed dynamically and unpredicably in he course of execuing he algorihm. This is he case when many recursive divide-and-conquer algorihms have o be used, including backrack search, game ree search and branch-and-bound compuaion. When all queue lenghs and produciviies (performance) of nodes are known, hen he bes conrol sraegy is a proporional disribuion of asks such ha q 1 /θ 1 = q 2 /θ 2 = = q m /θ m. The proof of his resul is no difficul and could be found, for example, in Amelina e al. (2015). This conrol sraegy is called load balancing. The reasons menioned above allow us o reformulae he considering problem: he goal is o mainain he balanced (equal) load across he nework. Assume, ha he following condiions are saisfied A1: Graph G A is undireced, and i has a spanning ree. A2: θ i θ min > 0, i N, = 0, 1,.... (Noe, if Assumpion A1 is saisfied hen 0 < Re(λ 2 (A)) (see Lewis e al. (2014))). If we ake x i = q/θ i i as a sae of agen i of considered dynamic nework a ime insans = 0, 1..., hen he conrol goal of achieving consensus in nework will correspond o he opimal redisribuion of asks among agens (see Amelina e al. (2015)). Under his noaion, he dynamics of each agen can be rewrien as x i +1 = x i + f i + ũ i, (5) where f i = z i /θ i 1, and ũ i = ū i /θ i, i N are normalized conrol acions. We can rewrie Equaion (5) in he vecor form x +1 = x + f + u, (6) where m-vecors x, f, and u consis of corresponding elemens x 1,..., x m, f 1,..., f m, and ũ 1,..., ũ m. If undireced graph G A has a spanning ree, he load balancing problem can be reformulaed as a minimizaion problem of a Laplacian poenial associaed wih graph G A (see Olfai-Saber and Murray (2004)) 947
3 June 24-26, Sain Peersburg, Russia Φ (x ) = 1 2 n j=1 subjec o m a i,j (x j x i ) 2 min x, (7) m m x i θ i = q 1, i (8) since Φ (x ) = 0 for he case x 1 = x 2 =... = x m and Φ (x ) > 0 for all oher cases. Is is also menioned in Olfai- Saber and Murray (2004) ha local voing proocol (see, e.g., Amelina e al. (2015)) is equivalen o gradien descen for Laplacian poenial. Linear consrain (8) is naural for problems of asks redisribuion because we canno loss he asks during a redisribuion process. To solve he problem (7),(8) we could use he algorihm and resul from Granichin (2015). 4. MEAN-RISK OPTIMIZATION PROBLEM UNDER LINEAR CONSTRAINS Consider a se of differeniable funcions {f w (θ)} w W, f w (θ) : R m R, le x 1, x 2,... be he se of observaion poins chosen by experimener. For each = 1, 2,... we ge measuremens y 1, y 2,... of f w ( ) wih addiive exernal noise v y = f w (x ) + v, (9) where {w } is an unconrollable sequence, w W. Le (Ω, F, P ) be he underlying probabiliy space, and le F 1 be he σ-algebra of all probabilisic evens occurred before = 1, 2,.... The problem is o find opimal θ ha minimizes mean-risk funcional F (θ) = E F 1 f w (θ) min (10) θ subjec o linear consrains H θ = q 1 (11) wih marices H of dimension k m and vecors q 1 R k, 0 k < m (wih k = 0 i is assumed ha here is no consrains). Hereandafer E is a symbol for mean value and E F 1 is a symbol for condiional mahemaical expecaion wih respec o F 1,, is a scalar produc of wo vecors, is an Euclidean norm of a vecor. If rankh = k hen here exiss linear funcion h : R m R m k and is reverse funcion g : R m k R m such as x = g (h (x)), x M = {H x = q 1 }. We assume ha h ( ) could always be chosen. Le n, n = 1, 2,... be an observed sequence of independen random variables in R m k, called he simulaneous es perurbaion, wih Bernoulli disribuion which elemens equal ±1 wih probabiliies 1 2. Le us ake a fixed iniial vecor θ 0 R m and choose posiive numbers α and β. Consider he algorihm x ± n = g 2n 1± 1 (h 2 2n 1± 1 ( θ 2 2n 2 ) ± β n ), θ 2n 1 = g 2n 1 (h 2n 1 ( θ 2n 2 )), (12) y n θ + yn 2n = g 2n (h 2n ( θ 2n 1 ) α n ), 2β which is similar o one proposed in Granichin (2015) when H ( ) does no depend on. Nex, we assume he following abou f w (x), F (x) and uncerainies in he model: A3: Funcion F ( ) has unique minimum poin θ and z R m k z h (θ ), E F 1 z f w (g (z)) µ z h (θ ) 2 wih a consan µ > 0. A4: w W gradien z f w (g (z)) saisfies he Lipschiz condiion: z, z R d k z f w (z ) z f w (z ) M z z wih a consan M µ. A5: Vecor-gradien f ( ) is uniformly bounded in poin h (θ ): E f (h(θ )) c 1, E f (h(θ )) 2 c 2, E f (h(θ )), f 1 (h(θ 1)) c 2 (c 1 = c 2 = 0 if w is nonrandom, i.e. f w (x) = F (x)). A6: For n = 1, 2,..., a) n and w 2n 1, w 2n (if hey are random) do no depend on σ-algebra F 2n 2. b) If w 2n 1, w 2n are random hen random vecors n and elemens w 2n 1, w 2n are independen. c) he successive differences v n = v 2n v 2n 1 of observaion noises are bounded: v n c v <, or E v n 2 c 2 v, if a sequence {v } is random. d) If v n is random hen v n and vecor n are independen. A7: Marices H 2n 1 and H 2n (if hey are random) do no depend on σ-algebra F 2n 2. A8: The drif is bounded: h (θ θ 1)) δ θ <, or E h (θ θ 1) 2 δθ 2 and E h (θ θ 1) h(θ 1 θ 2) δθ 2, if a sequence {w } is random. The rae of drif is bounded in a such way ha z R d k : E F2n 2 φ n (z) 2 c 3 z h (θ2n 2) 2 + c 4, where φ n (x) = f w2n (x) f w2n 1 (x). Denoe κ = 2(µ αγ), b = 2βMc 3 (1 + 6αMc2 ) + δ θ(m + 2µ + 6αM 2 c 4 ), l = 2αc 2 (c 2 c v + 3(max 4 n 2β + c2 (c 2 + M 2 (δ θ +2βc ) 2 )))+2δ θ (4βMc 3 +Mδ θ +c 1 +3µδθ 2 ), where γ = 3c 2 (M 2 c 2 + c 3 2β ). The following Theorem shows he asympoically efficien mean-squared weak upper bound of esimaion residuals by algorihm (12). Theorem 1. If rankh = k, assumpions A3-A8 hold, and α is sufficienly small: α (0; µ/γ) if µ 2 > 2γ, or α (0; µ µ 2 2γ 2γ ) ( µ+ µ 2 2γ 2γ ; µ/γ) oherwise, hen he sequence of esimaes provided by he algorihm (12) has asympoically efficien mean-squared weak upper bound of esimaion residuals L = (b + b 2 + κ l)/κ, (13) i.e. ε > 0 N such ha n > N E θ 2n θ 2n 2 L + ε. Proof of Theorem 1 is slighly differen from he correspondence proof in Granichin (2015) since we consider more complicaed problem seing and addiional Assumpion A5. 5. TASK REDISTRIBUTION PROTOCOL Generally, o ensure load balancing across a nework (in order o increase he overall hroughpu of a sysem and o reduce 948
4 June 24-26, Sain Peersburg, Russia execuion ime) i is naurally o use he redisribuion proocol over ime. Minimum poin x of (7) vary over ime due o he sysem dynamics (6). Consider SPSA algorihm (12) wih nonvanishing sep-sizes for racking he changes x using h (x) = h(x) = col(x 1,..., x m 1 ) m g (z) = col(z 1,..., z m 1, qi 1 d 1 j=1 zj θ i ). We have iniial guess x 0 which is formed by q i 0/θ i 0, i = 1,..., m. Le α > 0 and β > 0 be fairly small sep-sizes. The ieraion sep consiss of Compue wo values y ± n = Φ 2n 1± 1 2 (g 2n 1± 1 2 (h( x 1) ± β )); (14) Compue quasigradien vecor n = n y + n y n 2β Ge new esimae x 2n 1 = g 2n 1 (h( x 2n 2 ); θ m ; (15) x 2n = g 2n (h( x 2n 2 ) α n ). (16) We canno use (16) in decenralized load balancing problem since each agen is able o use informaion abou is neighbors only. Consider he ih componen of he quasigradien vecor from (15). By virue (7) and (14) we have i = i Φ ( x 1 + β ) Φ ( x 1 β ) = 2β i 1 n n a k,j 4β k=1 j=1 ( (x j + β j x k β k ) 2 (x j β j x k + β k ) 2). By using he difference of squares (formula: a 2 b 2 = (a b)(a + b)), we derive n n i = i a k,j ( j k )(x j x k ) = k=1 j=1 (a i,j + a j,i )(1 i j )(x j x i )+ j N i n n a k,j ( j k )(x j x k ), i k i j i since ( i ) 2 = 1. Denoing η i = n k i n j i ak,j ( j k )(x j x k ) we ge i = 2 j N i a i,j (1 i j )(x j x i ) + i η i. Following by he SPSA ieraion sep (16) we could consider decenralized conrol proocol u i = α ( ) a i,j (1 i j θ i ) θ j y i,j y i,i, i N, (17) j N i where α > 0 is a sep-size of conrol proocol (17). For each i N he dynamics of he closed loop sysem wih proocol (17) is as follows x i +1 = x i + f i +α ( ) a i,j y i,j (1 i j ) θ j yi,i θ i. (18) j N i If we denoe marix B = [b i,j ], where b i,j = a i,j (1 i j ), hen properies of a similar conrol algorihm, called a local voing proocol, for a load balancing problem were sudied in Amelina e al. (2015). The common feaure is ha he conrol value of he local voing proocol for each agen was deermined by he weighed sum of differences beween he informaion abou he sae of he agen and he informaion abou is neighbors saes. However, he analysis in Amelina e al. (2015) was done only for he case of saisical noise (noise wih Gaussian disribuion) wih sandard zero-mean and bounded covariance properies. Here we consider he randomized modificaion (he special case) of he local voing proocol, which was inspired by SPSA mehods. Probably we could use weaker condiions abou observaion noise {v i,j } and disurbances f i if we assume he independence of simulaneous es perurbaion on noise and disurbances (see Granichin e al. (2015)). 5.1 Connecion o gossip algorihms For considered SPSA algorihm we used Bernoulli disribued simulaneous perurbaion bu in fac we can use differen wihou losing core properies of he SPSA. In Granichin e al. (2015) required condiions are presened (chaper 3, condiions (3.8)). One possible disribuion is as follows: 0, k i, j k n = ±1, w.p. 1 2, k = i or k = j 1, Assuring i n = j n k = i or k = j If we apply such o (18) here will be only wo nonzero coordinaes and whole sum conains single nonzero erm. Muliplier ( j n k n) is ±2 for nonzero erms and he sign doesn affec summary value. Wih ha, considered sochasic approximaion procedure can be described as ieraively pick random pair of indices and average values of corresponding coordinaes (value of β affecs how much he values are drawn o heir average). Such algorihms was previously sudied in Boyd e al. (2006) and are known as gossip algorihms. As we menioned in previous secions, SPSA algorihm works wih arbirary bounded noises. Wih ha, gossip algorihms should work wih arbirary noises as well if choice of a pair of indices does no correlae wih exernal noise. 6. ASYMPTOTIC MEAN SQUARE ε-consensus Definiion 1. n agens are said o achieve he asympoic mean square ε-consensus, if E x i 0 2 <, i N, and here exiss a sequence {x } such ha for all i N. lim E x i x 2 ε Assume ha he following assumpions are saisfied: A9: a) For each i, j = 1,..., m vecor and ξ i,j (if i is random) are independen. b) For each i = 1,..., m vecor and z i (if i is random) are independen. 949
5 June 24-26, Sain Peersburg, Russia c) For each i = 1,..., m vecor and θ i (if i is random) are independen. d) For all i, j N, = 1, 2,... observaion noise ξ i,j is bounded: ξ i,j c ξ <, or E(ξ i,j ) 2 c 2 ξ if ξi,j is random. e) For each i = 1,..., m z i is bounded: z i c z <, or E(z) i 2 c 2 z if z i is random. Le x 0 be he average of he iniial daa x 0 = 1 n x i 0 n and {x } is he rajecory of he averaged sysem x +1 = x + 1 n f i. (19) n Theorem 1 allows o derive he level of upper bound of he asympoic mean square ε-consensus: lim E x x 1 m 2 ε wih some ε which can be calculaed using Theorem 1resul. Here 1 m is m-vecor of ones, For considered case µ = Re(λ 2 (A)), M = 4d max (A). To verify he applicabiliy of Theorem 1 we need o check ha: 1: The funcion Φ ( ) is srongly convex on subspace X = {x R m : x T 1 m = x T 1 m }, i.e. i has a unique minimum poin x 1 m and (x x 1 m ) T Φ (x) µ x x 1 m 2, x X wih a consan µ = Re(λ 2 (A)) > 0. Calculaing he derivaives, we ge Φ (x) n n = a i,j (x j x i ) + a j,i (x i x j ). i j=1 j=1 Hence, gradien-vecor Φ (x) equals o 2L(A)x. The vecor 1 m is he righ eigenvecor of Laplacian marix L(A) and corresponding o he zero eigenvalue: L(A)1 m = 0. Sums of all elemens in rows of marix L(A) is equal o zero and, moreover, all he diagonal elemens are posiive and equal o he absolue value of he sum of all oher elemens in he row. By virue Assumpion A1 marix A has a spanning ree. By Lemma 2.10 from Ren and Beard (2008) he rank of marix L(A) equals m 1. Hence we can derive (x x 1 m ) T Φ (x) = 2(x x 1 m ) T L(A)x = 2(x x 1 m ) T L(A)(x x 1 m ) Re(λ 2 (A)) x x 1 m 2, 2: By using Gershgorin crieria (see Lewis e al. (2014)), we ge ha he gradien Φ (x) saisfies he Lipschiz condiion: x, x R m Φ (x ) Φ (x ) = 2 L(A)(x x ) 4d max (A) x x wih a consan M = 4d max (A) µ = Re(λ 2 (A)). 3: By virue Assumpion A9a he vecor does no depend on observaion noise and drif. 4: By virue Assumpion A9d he observaion noise saisfies: ( ) a i,j ξ i,j (1 i j ) θ j ξi,i θ i j N i 4d max (A)c v /θ min <. 5: By virue Assumpion A9e he drif is bounded: x 1 m x 11 m = 1 m f i m m max{1, c z /θ min 1} n = δ x <. 7. SIMULATION RESULTS To illusrae he heoreical resuls we consider he decenralized compuer nework of m = 50 compuing nodes. We will show ha he proposed randomized conrol algorihm (17) provides load balancing of he nework similar o he one presened in Fig. 1. Fig. 1. The nework opology. The nework opology is a ring wih chords which are randomly chosen by he following rule for every node: (1) simulae a number of added chords by a Poisson disribuion wih mean value m/2 (2) randomly selec nodes ha aach o he curren (he number of such unis is equal o he value obained in sep 1) We generae he iniial produciviies θ 1, θ 2,..., θ m randomly by he uniform disribuion over he inerval (10; 50). We assume ha produciviy measure in our case is he number of available jobs in ime insan = 0, 1,..., he produciviies do no change over ime and θ i 0 i. The asks are divided ino wo ses: regular and burs. The firs one is served on each ac o a randomly chosen node and he second one a any given ime. During sysem operaion we will be adding regular asks from he inerval (12; 100) and burs asks from (10000; 25000). Fig. 2 shows he dependence of algorihm convergence rae on choosing of coefficien α. In Fig. 3, we can see he sysem of m = 50 nodes operaing in nonsaionary case wih he conrol proocol (17). Each line indicaes how he load x i evolves over ime. For clariy, he char displays 3 maximum and 3 minimum values. These lines also show how he sysem evolves o reach load-balancing or consensus. We can see ha even when he new burs ask se is received during he sysem work, i does no affec he qualiy of load balancing. During he simulaion we have se he coefficien α = 0.007, which is he mos suiable value for he curren opology and chosen parameers (see Fig. 2). In addiion 950
6 June 24-26, Sain Peersburg, Russia Fig. 2. Rae of convergence based on α. o he obained resuls, i is planned o sudy he possibiliy of SPSA applicaion for racking he opimal value of α. Fig. 3. Perfomance of he sysem wih m = 50 nodes x i for he nonsaionary case. 8. CONCLUSION In his paper he problem of load balancing in a muli-agen sysem under unknown bu bounded disurbances was examined. To solve he load balancing problem he new randomized local voing proocol wih nonvanishing sep-size was proposed. Condiions for achieving an approximae consensus (balance of he nework load) were obained. To illusrae he heoreical resuls we presened he simulaions for he compuing nework. REFERENCES Amelin, K., Amelina, N., Granichin, O., Granichina, O., and Andrievsky, B. (2013). Randomized algorihm for uavs group fligh opimizaion. 11h IFAC Inernaional Workshop on Adapaion and Learning in Conrol and Signal Processing, Amelina, N. and Fradkov, A. (2012). Approximae consensus in he dynamic sochasic nework wih incomplee informaion and measuremen delays. Auomaion and Remoe Conrol, 73(11), Amelina, N., Fradkov, A., Jiang, Y., and Vergados, D. (2015). Approximae consensus in sochasic neworks wih applicaion o load balancing. IEEE Transacions on Informaion Theory, 61(4), Boyd, S., Ghosh, A., Prabhakar, B., and Shah, D. (2006). Randomized gossip algorihms. IEEE Transacions on Informaion Theory, 52(6), Cheboarev, P.Y. and Agaev, R.P. (2009). Coordinaion in muliagen sysems and laplacian specra of digraphs. Auomaion and Remoe Conrol, 70(3), Granichin, O. and Amelina, N. (2015). Simulaneous perurbaion sochasic approximaion for racking under unknown bu bounded disurbances. IEEE Transacions on Auomaic Conrol, 60(5). Granichin, O., Skobelev, P., Lada, A., Mayorov, I., and Tsarev, A. (2012). Comparing adapive and non-adapive models of cargo ransporaion in muli-agen sysem for real ime ruck scheduling. Proceedings of he 4h Inernaional Join Conference on Compuaional Inelligence, Granichin, O., Volkovich, Z.V., and Toledano-Kiai, D. (2015). Randomized Algorihms in Auomaic Conrol and Daa Mining. Springer. Granichin, O. (1989). A sochasic recursive procedure wih correlaed noise in he observaion, ha employs rial perurbaions a he inpu. Vesnik Leningrad Universiy: Mah, 22(1), Granichin, O. (1992). Unknown funcion minimum poin esimaion under dependen noise. Problems of Informaion Transmission, 28(2), Granichin, O. (2015). Sochasic approximaion search algorihms wih randomizaion a he inpu. Auomaion and Remoe Conrol, 76(5), Huang, M. and Manon, J. (2009). Coordinaion and consensus of neworked agens wih noisy measuremens: sochasic algorihms and asympoic behavior. SIAM Journal on Conrol and Opimizaion, 48(1), Kar, S. and Moura, J.M. (2009). Disribued consensus algorihms in sensor neworks wih imperfec communicaion: Link failures and channel noise. IEEE Transacions on Signal Processing, 57(1), Lewis, F.L., Zhang, H., Hengser-Movric, K., and Das, A. (2014). Cooperaive conrol of muli-agen sysems: opimal and adapive design approaches. Springer Publishing Company, Incorporaed. Li, T. and Zhang, J. (2009). Mean square average-consensus under measuremen noises and fixed opologies: Necessary and sufficien condiions. Auomaica, 45(8), Olfai-Saber, R., Fax, J., and Murray, R. (2007). Consensus and cooperaion in neworked muli-agen sysems. Proceedings of he IEEE, 95(1), Olfai-Saber, R. and Murray, R. (2004). Consensus problems in neworks of agens wih swiching opology and ime-delays. IEEE Transacions on Auomaic Conrol, 49(9), Polyak, B.T. and Tsybakov, A.B. (1990). Opimal order of accuracy of search algorihms in sochasic opimizaion. Problemy Peredachi Informasii, 26(2), Rasrigin, L. (1963). The convergence of he random search mehod in he exremal conrol of a many-parameer sysem. Auomaion and Remoe Conrol, 24(10), Ren, W. and Beard, R. (2008). Disribued Consensus in Mulivehicle Cooperaive Conrol. Communicaions and Conrol Engineering. Springer. Ren, W., Beard, R., and Akins, E. (2007). Informaion consensus in mulivehicle cooperaive conrol. Conrol Sysems, IEEE, 27(2), Spall, J.C. (1992). Mulivariae sochasic approximaion using a simulaneous perurbaion gradien approximaion. IEEE Transacions on Auomaic Conrol, 37(3), Tsisiklis, J., Bersekas, D., and Ahans, M. (1986). Disribued asynchronous deerminisic and sochasic gradien opimizaion algorihms. IEEE Transacions on Auomaic Conrol, 31(9),
Network Discovery: An Estimation Based Approach
Nework Discovery: An Esimaion Based Approach Girish Chowdhary, Magnus Egersed, and Eric N. Johnson Absrac We consider he unaddressed problem of nework discovery, in which, an agen aemps o formulae an esimae
ANALYSIS AND COMPARISONS OF SOME SOLUTION CONCEPTS FOR STOCHASTIC PROGRAMMING PROBLEMS
ANALYSIS AND COMPARISONS OF SOME SOLUTION CONCEPTS FOR STOCHASTIC PROGRAMMING PROBLEMS R. Caballero, E. Cerdá, M. M. Muñoz and L. Rey () Deparmen of Applied Economics (Mahemaics), Universiy of Málaga,
Chapter 7. Response of First-Order RL and RC Circuits
Chaper 7. esponse of Firs-Order L and C Circuis 7.1. The Naural esponse of an L Circui 7.2. The Naural esponse of an C Circui 7.3. The ep esponse of L and C Circuis 7.4. A General oluion for ep and Naural
The Transport Equation
The Transpor Equaion Consider a fluid, flowing wih velociy, V, in a hin sraigh ube whose cross secion will be denoed by A. Suppose he fluid conains a conaminan whose concenraion a posiion a ime will be
Single-machine Scheduling with Periodic Maintenance and both Preemptive and. Non-preemptive jobs in Remanufacturing System 1
Absrac number: 05-0407 Single-machine Scheduling wih Periodic Mainenance and boh Preempive and Non-preempive jobs in Remanufacuring Sysem Liu Biyu hen Weida (School of Economics and Managemen Souheas Universiy
Real-time Particle Filters
Real-ime Paricle Filers Cody Kwok Dieer Fox Marina Meilă Dep. of Compuer Science & Engineering, Dep. of Saisics Universiy of Washingon Seale, WA 9895 ckwok,fox @cs.washingon.edu, [email protected] Absrac
DYNAMIC MODELS FOR VALUATION OF WRONGFUL DEATH PAYMENTS
DYNAMIC MODELS FOR VALUATION OF WRONGFUL DEATH PAYMENTS Hong Mao, Shanghai Second Polyechnic Universiy Krzyszof M. Osaszewski, Illinois Sae Universiy Youyu Zhang, Fudan Universiy ABSTRACT Liigaion, exper
adaptive control; stochastic systems; certainty equivalence principle; long-term
COMMUICATIOS I IFORMATIO AD SYSTEMS c 2006 Inernaional Press Vol. 6, o. 4, pp. 299-320, 2006 003 ADAPTIVE COTROL OF LIEAR TIME IVARIAT SYSTEMS: THE BET O THE BEST PRICIPLE S. BITTATI AD M. C. CAMPI Absrac.
Analogue and Digital Signal Processing. First Term Third Year CS Engineering By Dr Mukhtiar Ali Unar
Analogue and Digial Signal Processing Firs Term Third Year CS Engineering By Dr Mukhiar Ali Unar Recommended Books Haykin S. and Van Veen B.; Signals and Sysems, John Wiley& Sons Inc. ISBN: 0-7-380-7 Ifeachor
Distributed and Secure Computation of Convex Programs over a Network of Connected Processors
DCDIS CONFERENCE GUELPH, ONTARIO, CANADA, JULY 2005 1 Disribued and Secure Compuaion of Convex Programs over a Newor of Conneced Processors Michael J. Neely Universiy of Souhern California hp://www-rcf.usc.edu/
Optimal Investment and Consumption Decision of Family with Life Insurance
Opimal Invesmen and Consumpion Decision of Family wih Life Insurance Minsuk Kwak 1 2 Yong Hyun Shin 3 U Jin Choi 4 6h World Congress of he Bachelier Finance Sociey Torono, Canada June 25, 2010 1 Speaker
Monte Carlo Observer for a Stochastic Model of Bioreactors
Mone Carlo Observer for a Sochasic Model of Bioreacors Marc Joannides, Irène Larramendy Valverde, and Vivien Rossi 2 Insiu de Mahémaiques e Modélisaion de Monpellier (I3M UMR 549 CNRS Place Eugène Baaillon
Stochastic Optimal Control Problem for Life Insurance
Sochasic Opimal Conrol Problem for Life Insurance s. Basukh 1, D. Nyamsuren 2 1 Deparmen of Economics and Economerics, Insiue of Finance and Economics, Ulaanbaaar, Mongolia 2 School of Mahemaics, Mongolian
Chapter 8: Regression with Lagged Explanatory Variables
Chaper 8: Regression wih Lagged Explanaory Variables Time series daa: Y for =1,..,T End goal: Regression model relaing a dependen variable o explanaory variables. Wih ime series new issues arise: 1. One
TEMPORAL PATTERN IDENTIFICATION OF TIME SERIES DATA USING PATTERN WAVELETS AND GENETIC ALGORITHMS
TEMPORAL PATTERN IDENTIFICATION OF TIME SERIES DATA USING PATTERN WAVELETS AND GENETIC ALGORITHMS RICHARD J. POVINELLI AND XIN FENG Deparmen of Elecrical and Compuer Engineering Marquee Universiy, P.O.
Task is a schedulable entity, i.e., a thread
Real-Time Scheduling Sysem Model Task is a schedulable eniy, i.e., a hread Time consrains of periodic ask T: - s: saring poin - e: processing ime of T - d: deadline of T - p: period of T Periodic ask T
Random Walk in 1-D. 3 possible paths x vs n. -5 For our random walk, we assume the probabilities p,q do not depend on time (n) - stationary
Random Walk in -D Random walks appear in many cones: diffusion is a random walk process undersanding buffering, waiing imes, queuing more generally he heory of sochasic processes gambling choosing he bes
Journal Of Business & Economics Research September 2005 Volume 3, Number 9
Opion Pricing And Mone Carlo Simulaions George M. Jabbour, (Email: [email protected]), George Washingon Universiy Yi-Kang Liu, ([email protected]), George Washingon Universiy ABSTRACT The advanage of Mone Carlo
A Generalized Bivariate Ornstein-Uhlenbeck Model for Financial Assets
A Generalized Bivariae Ornsein-Uhlenbeck Model for Financial Asses Romy Krämer, Mahias Richer Technische Universiä Chemniz, Fakulä für Mahemaik, 917 Chemniz, Germany Absrac In his paper, we sudy mahemaical
ARCH 2013.1 Proceedings
Aricle from: ARCH 213.1 Proceedings Augus 1-4, 212 Ghislain Leveille, Emmanuel Hamel A renewal model for medical malpracice Ghislain Léveillé École d acuaria Universié Laval, Québec, Canada 47h ARC Conference
MTH6121 Introduction to Mathematical Finance Lesson 5
26 MTH6121 Inroducion o Mahemaical Finance Lesson 5 Conens 2.3 Brownian moion wih drif........................... 27 2.4 Geomeric Brownian moion........................... 28 2.5 Convergence of random
Multiprocessor Systems-on-Chips
Par of: Muliprocessor Sysems-on-Chips Edied by: Ahmed Amine Jerraya and Wayne Wolf Morgan Kaufmann Publishers, 2005 2 Modeling Shared Resources Conex swiching implies overhead. On a processing elemen,
Measuring macroeconomic volatility Applications to export revenue data, 1970-2005
FONDATION POUR LES ETUDES ET RERS LE DEVELOPPEMENT INTERNATIONAL Measuring macroeconomic volailiy Applicaions o expor revenue daa, 1970-005 by Joël Cariolle Policy brief no. 47 March 01 The FERDI is a
SPEC model selection algorithm for ARCH models: an options pricing evaluation framework
Applied Financial Economics Leers, 2008, 4, 419 423 SEC model selecion algorihm for ARCH models: an opions pricing evaluaion framework Savros Degiannakis a, * and Evdokia Xekalaki a,b a Deparmen of Saisics,
DETERMINISTIC INVENTORY MODEL FOR ITEMS WITH TIME VARYING DEMAND, WEIBULL DISTRIBUTION DETERIORATION AND SHORTAGES KUN-SHAN WU
Yugoslav Journal of Operaions Research 2 (22), Number, 6-7 DEERMINISIC INVENORY MODEL FOR IEMS WIH IME VARYING DEMAND, WEIBULL DISRIBUION DEERIORAION AND SHORAGES KUN-SHAN WU Deparmen of Bussines Adminisraion
How To Predict A Person'S Behavior
Informaion Theoreic Approaches for Predicive Models: Resuls and Analysis Monica Dinculescu Supervised by Doina Precup Absrac Learning he inernal represenaion of parially observable environmens has proven
Optimal Stock Selling/Buying Strategy with reference to the Ultimate Average
Opimal Sock Selling/Buying Sraegy wih reference o he Ulimae Average Min Dai Dep of Mah, Naional Universiy of Singapore, Singapore Yifei Zhong Dep of Mah, Naional Universiy of Singapore, Singapore July
Inductance and Transient Circuits
Chaper H Inducance and Transien Circuis Blinn College - Physics 2426 - Terry Honan As a consequence of Faraday's law a changing curren hrough one coil induces an EMF in anoher coil; his is known as muual
Duration and Convexity ( ) 20 = Bond B has a maturity of 5 years and also has a required rate of return of 10%. Its price is $613.
Graduae School of Business Adminisraion Universiy of Virginia UVA-F-38 Duraion and Convexiy he price of a bond is a funcion of he promised paymens and he marke required rae of reurn. Since he promised
Time Series Analysis Using SAS R Part I The Augmented Dickey-Fuller (ADF) Test
ABSTRACT Time Series Analysis Using SAS R Par I The Augmened Dickey-Fuller (ADF) Tes By Ismail E. Mohamed The purpose of his series of aricles is o discuss SAS programming echniques specifically designed
The naive method discussed in Lecture 1 uses the most recent observations to forecast future values. That is, Y ˆ t + 1
Business Condiions & Forecasing Exponenial Smoohing LECTURE 2 MOVING AVERAGES AND EXPONENTIAL SMOOTHING OVERVIEW This lecure inroduces ime-series smoohing forecasing mehods. Various models are discussed,
SELF-EVALUATION FOR VIDEO TRACKING SYSTEMS
SELF-EVALUATION FOR VIDEO TRACKING SYSTEMS Hao Wu and Qinfen Zheng Cenre for Auomaion Research Dep. of Elecrical and Compuer Engineering Universiy of Maryland, College Park, MD-20742 {wh2003, qinfen}@cfar.umd.edu
Newton s Laws of Motion
Newon s Laws of Moion MS4414 Theoreical Mechanics Firs Law velociy. In he absence of exernal forces, a body moves in a sraigh line wih consan F = 0 = v = cons. Khan Academy Newon I. Second Law body. The
Technical Appendix to Risk, Return, and Dividends
Technical Appendix o Risk, Reurn, and Dividends Andrew Ang Columbia Universiy and NBER Jun Liu UC San Diego This Version: 28 Augus, 2006 Columbia Business School, 3022 Broadway 805 Uris, New York NY 10027,
ON THE PRICING OF EQUITY-LINKED LIFE INSURANCE CONTRACTS IN GAUSSIAN FINANCIAL ENVIRONMENT
Teor Imov r.amaem.sais. Theor. Probabiliy and Mah. Sais. Vip. 7, 24 No. 7, 25, Pages 15 111 S 94-9(5)634-4 Aricle elecronically published on Augus 12, 25 ON THE PRICING OF EQUITY-LINKED LIFE INSURANCE
The Application of Multi Shifts and Break Windows in Employees Scheduling
The Applicaion of Muli Shifs and Brea Windows in Employees Scheduling Evy Herowai Indusrial Engineering Deparmen, Universiy of Surabaya, Indonesia Absrac. One mehod for increasing company s performance
Distributing Human Resources among Software Development Projects 1
Disribuing Human Resources among Sofware Developmen Proecs Macario Polo, María Dolores Maeos, Mario Piaini and rancisco Ruiz Summary This paper presens a mehod for esimaing he disribuion of human resources
Term Structure of Prices of Asian Options
Term Srucure of Prices of Asian Opions Jirô Akahori, Tsuomu Mikami, Kenji Yasuomi and Teruo Yokoa Dep. of Mahemaical Sciences, Risumeikan Universiy 1-1-1 Nojihigashi, Kusasu, Shiga 525-8577, Japan E-mail:
LIFE INSURANCE WITH STOCHASTIC INTEREST RATE. L. Noviyanti a, M. Syamsuddin b
LIFE ISURACE WITH STOCHASTIC ITEREST RATE L. oviyani a, M. Syamsuddin b a Deparmen of Saisics, Universias Padjadjaran, Bandung, Indonesia b Deparmen of Mahemaics, Insiu Teknologi Bandung, Indonesia Absrac.
Dependent Interest and Transition Rates in Life Insurance
Dependen Ineres and ransiion Raes in Life Insurance Krisian Buchard Universiy of Copenhagen and PFA Pension January 28, 2013 Absrac In order o find marke consisen bes esimaes of life insurance liabiliies
DDoS Attacks Detection Model and its Application
DDoS Aacks Deecion Model and is Applicaion 1, MUHAI LI, 1 MING LI, XIUYING JIANG 1 School of Informaion Science & Technology Eas China Normal Universiy No. 500, Dong-Chuan Road, Shanghai 0041, PR. China
AP Calculus AB 2013 Scoring Guidelines
AP Calculus AB 1 Scoring Guidelines The College Board The College Board is a mission-driven no-for-profi organizaion ha connecs sudens o college success and opporuniy. Founded in 19, he College Board was
Stock Trading with Recurrent Reinforcement Learning (RRL) CS229 Application Project Gabriel Molina, SUID 5055783
Sock raing wih Recurren Reinforcemen Learning (RRL) CS9 Applicaion Projec Gabriel Molina, SUID 555783 I. INRODUCION One relaively new approach o financial raing is o use machine learning algorihms o preic
Research on Inventory Sharing and Pricing Strategy of Multichannel Retailer with Channel Preference in Internet Environment
Vol. 7, No. 6 (04), pp. 365-374 hp://dx.doi.org/0.457/ijhi.04.7.6.3 Research on Invenory Sharing and Pricing Sraegy of Mulichannel Reailer wih Channel Preference in Inerne Environmen Hanzong Li College
CHARGE AND DISCHARGE OF A CAPACITOR
REFERENCES RC Circuis: Elecrical Insrumens: Mos Inroducory Physics exs (e.g. A. Halliday and Resnick, Physics ; M. Sernheim and J. Kane, General Physics.) This Laboraory Manual: Commonly Used Insrumens:
Appendix A: Area. 1 Find the radius of a circle that has circumference 12 inches.
Appendi A: Area worked-ou s o Odd-Numbered Eercises Do no read hese worked-ou s before aemping o do he eercises ourself. Oherwise ou ma mimic he echniques shown here wihou undersanding he ideas. Bes wa
Task-Execution Scheduling Schemes for Network Measurement and Monitoring
Task-Execuion Scheduling Schemes for Nework Measuremen and Monioring Zhen Qin, Robero Rojas-Cessa, and Nirwan Ansari Deparmen of Elecrical and Compuer Engineering New Jersey Insiue of Technology Universiy
A Scalable and Lightweight QoS Monitoring Technique Combining Passive and Active Approaches
A Scalable and Lighweigh QoS Monioring Technique Combining Passive and Acive Approaches On he Mahemaical Formulaion of CoMPACT Monior Masai Aida, Naoo Miyoshi and Keisue Ishibashi NTT Informaion Sharing
An Online Learning-based Framework for Tracking
An Online Learning-based Framework for Tracking Kamalika Chaudhuri Compuer Science and Engineering Universiy of California, San Diego La Jolla, CA 9293 Yoav Freund Compuer Science and Engineering Universiy
PRECISE positioning/tracking control is being studied
Design of High Accuracy Tracking Sysems wih H Preview Conrol Anonio Moran Cardenas, Javier G. Rázuri, Isis Bone, Rahim Rahmani, and David Sundgren Absrac Posiioning and racking conrol sysems are an imporan
Module 4. Single-phase AC circuits. Version 2 EE IIT, Kharagpur
Module 4 Single-phase A circuis ersion EE T, Kharagpur esson 5 Soluion of urren in A Series and Parallel ircuis ersion EE T, Kharagpur n he las lesson, wo poins were described:. How o solve for he impedance,
A Resource Management Strategy to Support VoIP across Ad hoc IEEE 802.11 Networks
A Resource Managemen Sraegy o Suppor VoIP across Ad hoc IEEE 8.11 Neworks Janusz Romanik Radiocommunicaions Deparmen Miliary Communicaions Insiue Zegrze, Poland [email protected] Pior Gajewski, Jacek
Q-SAC: Toward QoS Optimized Service Automatic Composition *
Q-SAC: Toward QoS Opimized Service Auomaic Composiion * Hanhua Chen, Hai Jin, Xiaoming Ning, Zhipeng Lü Cluser and Grid Compuing Lab Huazhong Universiy of Science and Technology, Wuhan, 4374, China Email:
A Natural Feature-Based 3D Object Tracking Method for Wearable Augmented Reality
A Naural Feaure-Based 3D Objec Tracking Mehod for Wearable Augmened Realiy Takashi Okuma Columbia Universiy / AIST Email: [email protected] Takeshi Kuraa Universiy of Washingon / AIST Email: [email protected]
Improvement of a TCP Incast Avoidance Method for Data Center Networks
Improvemen of a Incas Avoidance Mehod for Daa Cener Neworks Kazuoshi Kajia, Shigeyuki Osada, Yukinobu Fukushima and Tokumi Yokohira The Graduae School of Naural Science and Technology, Okayama Universiy
On the degrees of irreducible factors of higher order Bernoulli polynomials
ACTA ARITHMETICA LXII.4 (1992 On he degrees of irreducible facors of higher order Bernoulli polynomials by Arnold Adelberg (Grinnell, Ia. 1. Inroducion. In his paper, we generalize he curren resuls on
Individual Health Insurance April 30, 2008 Pages 167-170
Individual Healh Insurance April 30, 2008 Pages 167-170 We have received feedback ha his secion of he e is confusing because some of he defined noaion is inconsisen wih comparable life insurance reserve
Economics Honors Exam 2008 Solutions Question 5
Economics Honors Exam 2008 Soluions Quesion 5 (a) (2 poins) Oupu can be decomposed as Y = C + I + G. And we can solve for i by subsiuing in equaions given in he quesion, Y = C + I + G = c 0 + c Y D + I
Particle Filtering for Geometric Active Contours with Application to Tracking Moving and Deforming Objects
Paricle Filering for Geomeric Acive Conours wih Applicaion o Tracking Moving and Deforming Objecs Yogesh Rahi Namraa Vaswani Allen Tannenbaum Anhony Yezzi Georgia Insiue of Technology School of Elecrical
COMPARISON OF AIR TRAVEL DEMAND FORECASTING METHODS
COMPARISON OF AIR RAVE DEMAND FORECASING MEHODS Ružica Škurla Babić, M.Sc. Ivan Grgurević, B.Eng. Universiy of Zagreb Faculy of ranspor and raffic Sciences Vukelićeva 4, HR- Zagreb, Croaia [email protected],
Mathematics in Pharmacokinetics What and Why (A second attempt to make it clearer)
Mahemaics in Pharmacokineics Wha and Why (A second aemp o make i clearer) We have used equaions for concenraion () as a funcion of ime (). We will coninue o use hese equaions since he plasma concenraions
Hotel Room Demand Forecasting via Observed Reservation Information
Proceedings of he Asia Pacific Indusrial Engineering & Managemen Sysems Conference 0 V. Kachivichyanuul, H.T. Luong, and R. Piaaso Eds. Hoel Room Demand Forecasing via Observed Reservaion Informaion aragain
Efficient One-time Signature Schemes for Stream Authentication *
JOURNAL OF INFORMATION SCIENCE AND ENGINEERING, 611-64 (006) Efficien One-ime Signaure Schemes for Sream Auhenicaion * YONGSU PARK AND YOOKUN CHO + College of Informaion and Communicaions Hanyang Universiy
PATHWISE PROPERTIES AND PERFORMANCE BOUNDS FOR A PERISHABLE INVENTORY SYSTEM
PATHWISE PROPERTIES AND PERFORMANCE BOUNDS FOR A PERISHABLE INVENTORY SYSTEM WILLIAM L. COOPER Deparmen of Mechanical Engineering, Universiy of Minnesoa, 111 Church Sree S.E., Minneapolis, MN 55455 [email protected]
AP Calculus AB 2010 Scoring Guidelines
AP Calculus AB 1 Scoring Guidelines The College Board The College Board is a no-for-profi membership associaion whose mission is o connec sudens o college success and opporuniy. Founded in 1, he College
11/6/2013. Chapter 14: Dynamic AD-AS. Introduction. Introduction. Keeping track of time. The model s elements
Inroducion Chaper 14: Dynamic D-S dynamic model of aggregae and aggregae supply gives us more insigh ino how he economy works in he shor run. I is a simplified version of a DSGE model, used in cuing-edge
USE OF EDUCATION TECHNOLOGY IN ENGLISH CLASSES
USE OF EDUCATION TECHNOLOGY IN ENGLISH CLASSES Mehme Nuri GÖMLEKSİZ Absrac Using educaion echnology in classes helps eachers realize a beer and more effecive learning. In his sudy 150 English eachers were
Large Scale Online Learning.
Large Scale Online Learning. Léon Boou NEC Labs America Princeon NJ 08540 [email protected] Yann Le Cun NEC Labs America Princeon NJ 08540 [email protected] Absrac We consider siuaions where raining daa is abundan
Answer, Key Homework 2 David McIntyre 45123 Mar 25, 2004 1
Answer, Key Homework 2 Daid McInyre 4123 Mar 2, 2004 1 This prin-ou should hae 1 quesions. Muliple-choice quesions may coninue on he ne column or page find all choices before making your selecion. The
Option Put-Call Parity Relations When the Underlying Security Pays Dividends
Inernaional Journal of Business and conomics, 26, Vol. 5, No. 3, 225-23 Opion Pu-all Pariy Relaions When he Underlying Securiy Pays Dividends Weiyu Guo Deparmen of Finance, Universiy of Nebraska Omaha,
Photovoltaic Power Control Using MPPT and Boost Converter
23 Phoovolaic Power Conrol Using MPP and Boos Converer A.Aou, A.Massoum and M.Saidi Absrac he sudies on he phoovolaic sysem are exensively increasing because of a large, secure, essenially exhausible and
1. y 5y + 6y = 2e t Solution: Characteristic equation is r 2 5r +6 = 0, therefore r 1 = 2, r 2 = 3, and y 1 (t) = e 2t,
Homework6 Soluions.7 In Problem hrough 4 use he mehod of variaion of parameers o find a paricular soluion of he given differenial equaion. Then check your answer by using he mehod of undeermined coeffiens..
Feasibility of Quantum Genetic Algorithm in Optimizing Construction Scheduling
Feasibiliy of Quanum Geneic Algorihm in Opimizing Consrucion Scheduling Maser Thesis Baihui Song JUNE 2013 Commiee members: Prof.dr.ir. M.J.C.M. Herogh Dr. M. Blaauboer Dr. ir. H.K.M. van de Ruienbeek
Model-Based Monitoring in Large-Scale Distributed Systems
Model-Based Monioring in Large-Scale Disribued Sysems Diploma Thesis Carsen Reimann Chemniz Universiy of Technology Faculy of Compuer Science Operaing Sysem Group Advisors: Prof. Dr. Winfried Kalfa Dr.
INTEREST RATE FUTURES AND THEIR OPTIONS: SOME PRICING APPROACHES
INTEREST RATE FUTURES AND THEIR OPTIONS: SOME PRICING APPROACHES OPENGAMMA QUANTITATIVE RESEARCH Absrac. Exchange-raded ineres rae fuures and heir opions are described. The fuure opions include hose paying
APPLICATION OF THE KALMAN FILTER FOR ESTIMATING CONTINUOUS TIME TERM STRUCTURE MODELS: THE CASE OF UK AND GERMANY. January, 2005
APPLICATION OF THE KALMAN FILTER FOR ESTIMATING CONTINUOUS TIME TERM STRUCTURE MODELS: THE CASE OF UK AND GERMANY Somnah Chaeree* Deparmen of Economics Universiy of Glasgow January, 2005 Absrac The purpose
Towards Intrusion Detection in Wireless Sensor Networks
Towards Inrusion Deecion in Wireless Sensor Neworks Kroniris Ioannis, Tassos Dimiriou and Felix C. Freiling Ahens Informaion Technology, 19002 Peania, Ahens, Greece Email: {ikro,dim}@ai.edu.gr Deparmen
An Agent-based Bayesian Forecasting Model for Enhanced Network Security
An Agen-based Forecasing Model for Enhanced Nework Securiy J. PIKOULAS, W.J. BUCHANAN, Napier Universiy, Edinburgh, UK. M. MANNION, Glasgow Caledonian Universiy, Glasgow, UK. K. TRIANTAFYLLOPOULOS, Universiy
Dynamic programming models and algorithms for the mutual fund cash balance problem
Submied o Managemen Science manuscrip Dynamic programming models and algorihms for he muual fund cash balance problem Juliana Nascimeno Deparmen of Operaions Research and Financial Engineering, Princeon
4 Convolution. Recommended Problems. x2[n] 1 2[n]
4 Convoluion Recommended Problems P4.1 This problem is a simple example of he use of superposiion. Suppose ha a discree-ime linear sysem has oupus y[n] for he given inpus x[n] as shown in Figure P4.1-1.
PROFIT TEST MODELLING IN LIFE ASSURANCE USING SPREADSHEETS PART ONE
Profi Tes Modelling in Life Assurance Using Spreadshees PROFIT TEST MODELLING IN LIFE ASSURANCE USING SPREADSHEETS PART ONE Erik Alm Peer Millingon 2004 Profi Tes Modelling in Life Assurance Using Spreadshees
