Optimized Base-Station Cache Allocation for Cloud Radio Access Network with Multicast Backhaul

Size: px
Start display at page:

Download "Optimized Base-Station Cache Allocation for Cloud Radio Access Network with Multicast Backhaul"

Transcription

1 Optimized Base-Station Cache Aocation for Coud Radio Access Network with Muticast Backhau Binbin Dai, Student Member, IEEE, Ya-Feng Liu, Member, IEEE, and Wei Yu, Feow, IEEE arxiv: v [cs.it] 28 Apr 208 Abstract The performance of coud radio access network C-RAN) is imited by the finite capacities of the backhau inks connecting the centraized processor CP) with the basestations BSs), especiay when the backhau is impemented in a wireess medium. This paper proposes the use of wireess muticast together with BS caching, where the BSs pre-store contents of popuar fies, to augment the backhau of C-RAN. For a downink C-RAN consisting of a singe custer of BSs and wireess backhau, this paper studies the optima cache size aocation strategy among the BSs and the optima muticast beamforming transmission strategy at the CP such that the user s requested messages are deivered from the CP to the BSs in the most efficient way. We first state a muticast backhau rate expression based on a joint cache-channe coding scheme, which impies that arger cache sizes shoud be aocated to the BSs with weaker channes. We then formuate a two-timescae joint cache size aocation and beamforming design probem, where the cache is optimized offine based on the ong-term channe statistica information, whie the beamformer is designed during the fie deivery phase based on the instantaneous channe state information. By everaging the sampe approximation method and the aternating direction method of mutipiers ADMM), we deveop efficient agorithms for optimizing cache size aocation among the BSs, and quantify how much more cache shoud be aocated to the weaker BSs. We further consider the case with mutipe fies having different popuarities and show that it is in genera not optima to entirey cache the most popuar fies first. Numerica resuts show considerabe performance improvement of the optimized cache size aocation scheme over the uniform aocation and other heuristic schemes. Index Terms Aternating direction method of mutipiers ADMM), base-station BS) caching, coud radio access network C-RAN), data-sharing strategy, muticasting, wireess backhau I. INTRODUCTION Coud radio access network C-RAN) has been recognized as one of the enabing technoogies to meet the ever-increasing demand for higher data rates for the next generation 5G) wireess networks [2] [4]. In C-RAN, the base-stations BSs) Manuscript submitted on December 0, 207; revised on Apri, 208; accepted on Apri 8, 208. The materias in this paper have been presented in part at the IEEE Internationa Conference on Acoustics, Speech and Signa Processing ICASSP), Cagary, Canada, 208 []. This work was supported in part by the Natura Sciences and Engineering Research Counci NSERC) of Canada and in part by the Nationa Natura Science Foundation of China NSFC) grants 6749 and B. Dai and W. Yu are with The Edward S. Rogers Sr. Department of Eectrica and Computer Engineering, University of Toronto, Toronto, ON M5S 3G4, Canada e-mais: {bdai, weiyu}@comm.utoronto.ca). Y.-F. Liu is with the State Key Laboratory of Scientific and Engineering Computing, Institute of Computationa Mathematics and Scientific/Engineering Computing, Academy of Mathematics and Systems Science, Chinese Academy of Sciences, Beijing, 0090, China e-mai: yafiu@sec.cc.ac.cn). are connected to a centraized processor CP) through highspeed fronthau/backhau inks, which provide opportunities for cooperation among the BSs for inter-ce interference canceation. The performance of C-RAN depends cruciay on the capacity of the fronthau/backhau inks. The objective of this paper is to expore the benefit of utiizing caching at the BSs to augment the fronthau/backhau inks. There are two fundamentay different fronthauing strategies that enabe the cooperation of the BSs in C-RAN. In the data-sharing strategy [5] [8], the CP directy shares the user s messages with a custer of BSs, which subsequenty serve the user through cooperative beamforming. In the compression strategy [5], [9], the CP performs the beamforming operation and sends the compressed version of the anaog beamformed signa to the BSs. The reative advantage of the data-sharing strategy versus the compression strategy depends highy on the fronthau/backhau channe capacity [0], []. In genera, the compression strategy outperforms data-sharing when the fronthau/backhau capacity is moderatey high, in part because the data-sharing strategy reies on the backhau to carry each user s data mutipe times to mutipe cooperating BSs. Thus, the finite backhau capacity significanty imits the BS cooperation size. The capacity imitation in fronthau/backhau is especiay pertinent for sma-ce depoyment where high-speed fiber optic connections from the CP to the BSs may not be avaiabe and wireess backhauing may be the most feasibe engineering option. The purpose of this paper is to point out that under this scenario, the data-sharing strategy has a distinct edge in that it can take advantage of: i) the abiity of the CP to muticast user messages wireessy to mutipe BSs at the same time; and ii) the abiity of the BSs to cache user messages to further aeviate the backhau requirement. Note that the muticast opportunity in the wireess backhau and the caching opportunity at the BSs are ony avaiabe to faciitate the datasharing strategy in C-RAN, but not the compression strategy, as the atter invoves sending anaog compressed beamformed signas from the CP to the BSs, which are different for different BSs and are aso constanty changing according to the channe conditions, so are impossibe to cache. This paper considers a downink C-RAN in which the CP utiizes mutipe antennas to muticast user messages to a singe custer of BSs using the data-sharing strategy, whie the BSs pre-store fractions of popuar contents during the offpeak time and request the rest of the fies from the CP using coded deivery via the noisy wireess backhau channe. Given a tota cache constraint, we investigate the optima cache size aocation strategy across the BSs and the optima muticast

2 2 beamforming transmission strategy at the CP so that the fie requests can be deivered most efficienty from the CP to the BSs. It is important to emphasize that the optimizations of the BS cache size aocation and the beamforming strategy at the CP occur in different timescaes. Whie the beamformer can dynamicay adapt to the instantaneous channe reaization, the cache size is optimized ony at the cache aocation phase and can ony adapt to the ong-term statistics of the backhau channe. This paper proposes a sampe approximation approach to sove the above two-timescae optimization probem. The optima cache size aocation considers the ong-term channe statistics in aocating arger cache sizes to the BSs with weaker backhau channes, whie accounting for the potentia effect of beamforming. It aso considers the difference in fie popuarities in caching arger portions of more popuar fies. A. Reated Work Whie caching has been extensivey used at the edge of Internet, the idea of coded caching that takes advantage of muticast opportunity has recenty attracted extensive research interests due to the pioneering work of [2], which uses the network coding method to simutaneousy deiver mutipe fies through a common noiseess channe to mutipe receivers, each caching different parts of the fies. This paper studies a different scenario in which the same content is requested by mutipe receivers BSs), hence no network coding is needed and the coded muticasting in this paper refers to channe coding across mutipe wireess backhau channes with different channe conditions between the CP transmitter and the BS receivers. C-RAN with BS caching has been previousy considered in [3] [6], but most previous works assume fixed cache aocation among the BSs. More specificay, [3] and [4] examine how BS caching heps in reducing both backhau capacity consumption and BS power consumption under given users quaity-of-service constraints; [5] studies how BS caching changes the way that backhau is utiized and proposes a simiar scheme as in [7] that combines the data-sharing strategy and the compression strategy to improve the spectra efficiency of the downink C-RAN. This paper differs from the above works in focusing on how to optimay aocate the cache sizes among the BSs and design muticast beamformers at the CP to improve the efficiency of sharing user s requested fies via the wireess backhau channe. Previous works on caching strategy optimization rey on the assumptions of either simpified networks [8], [9] or Poisson distributed networks [20] [22] that are reasonabe in a network with a arge number of BSs and users, and focus on anayzing how BS caching heps in improving the performance of the BS-to-user ayer. This paper instead considers a C-RAN with a singe custer of BSs and investigates how BS caching improves the efficiency of fie deivery between the coud and the BSs ayer. This paper is motivated by [23] which shows from an information-theoretica perspective the advantage of aocating different cache sizes to different BSs depending on their channe conditions. In addition, [23] proposes a joint cachechanne coding scheme that optimay utiizes the caches at the BSs in a broadcast erasure channe, which is further generaized to the degraded broadcast channe in [24]. We take the findings in [23] one step further by considering the effect of mutipe-antenna beamforming in a downink C- RAN backhau network. We aso extend [23] to the case of mutipe fies with different popuarities and demonstrate that the optima caching strategy aso highy depends on the fie popuarities. B. Main Contributions This paper considers the joint optimization of BS cache size aocation and muticast beamformer at the CP in two timescaes for a downink C-RAN with a singe BS custer under the data-sharing strategy. The main contributions of this paper are summarized as foows: Probem Formuation: We derive a new muticast backhau rate expression with BS caching based on the joint cache-channe coding scheme of [23]. We then formuate two new cache size aocation probems of minimizing the expected fie downoading time and maximizing the expected fie downoading rate subject to the tota cache size constraint. The cache size aocation is optimized offine and is fixed during the fie deivery phase, whie the transmit beamformers are adapted to the rea-time channe reaization. Agorithms: We propose efficient agorithms for soving the formuated cache size aocation probems. More specificay, to dea with the intractabiity of taking expectation over the channe reaizations in the objective functions, we approximate the expectation via samping [25]. Note that the sampe size generay needs to be arge in order to guarantee the approximation accuracy. We further propose to sove the sampe approximation probem using the successive inear approximation technique and the aternating direction method of mutipiers ADMM) agorithm [26], which decomposes the potentiay argescae probem due to the arge sampe size) into many sma-scae probems on each sampe. Engineering Insight: We quantify how much cache shoud be aocated among the BSs in a practica C-RAN setup, and show that, as compared to the uniform and proportiona cache size aocation schemes, the proposed scheme aocates aggressivey arger cache sizes to the fies with higher popuarities, and for each fie the proposed scheme aocates aggressivey arger cache sizes to the BSs with weaker backhau channes. C. Paper Organization and Notations The remainder of this paper is organized as foows. Section II introduces the considered system mode for C-RAN. We derive the backhau muticast rate with BS caching in Section III and state the probem setup in Section IV. Sections V and VI focus on the proposed cache size aocation schemes for the singe fie case and the mutipe fies case,

3 3 repacements Centra Processor Wireess Backhau Cache Cache Cache BS BS 2 BS 3 User Fig.. Downink C-RAN with BS caching, where each BS is equipped with a oca storage unit that caches some contents of user s requested fies. respectivey. Simuation resuts are provided in Section VII. Concusions are drawn in Section VIII. Throughout this paper, ower-case etters e.g. x) and owercase bod etters e.g. x) denote scaars and coumn vectors, respectivey. We use C to denote the compex domain. The transpose and conjugate transpose of a vector are denoted as ) T and ) H, respectivey. The expectation of a random variabe is denoted as E[ ]. Caigraphy etters are used to denote sets. II. SYSTEM MODEL Consider a downink C-RAN mode in Fig. consisting of BSs connected to a coud-based CP through shared wireess backhau. The coud empoys a data-sharing strategy which deivers each user s intended message to a predefined custer of BSs and the BS custer subsequenty serves the user through cooperative beamforming. The capacities of the backhau is a significant imiting factor to the performance of the C-RAN [5], [8]. To aeviate the backhau requirement, this paper considers the scenario where each BS is equipped with a oca cache, as shown in Fig., that can pre-store a subset of the fies during off-peak traffic time in order to reduce the peak time backhau traffic. For simpicity, we consider a network consisting of a singe custer of L cooperative BSs, i.e., a the BSs in the network beamform cooperativey to serve each user. In this case, the user s intended message needs to be made avaiabe at a BSs in order to aow for cooperation. We assume that the CP has access to a the fies and deivers the user s requested fie to the BSs through muticast beamforming over the wireess backhau channe. The backhau connecting the CP with the BSs is impemented in a shared wireess medium, assumed to foow a bock-fading channe mode. We assume that the CP is equipped with M transmit antennas, whie a the BSs are equipped with a singe antenna. We denote the channe vector between the CP and the BS L := {,2,...,L}, as h C M, which remains constant within a coherent bock but changes independenty and identicay according to some distribution in different coherent bocks. The received signa at BS can be written as y = h H x+z, ) where x C M is the transmit signa of the CP transmitter, y C is the received signa at BS, and z CN 0, ) is the background noise at BS obeying the compex Gaussian distribution with zero mean and variance. Each BS is equipped with a oca cache of size C that can pre-store some contents of the fie. This paper addresses two questions: Given fixed oca cache sizes and fixed cached contents, at a fast timescae, how shoud the transmit beamforming strategy be designed as function of the instantaneous reaization of the wireess channe in order to most efficienty deivery a common user message to a the BSs? At a sow timescae, how shoud the contents be cached and how shoud the cache sizes be aocated across the BSs so that the expected deivery performance across many channe reaizations is optimized? The answers to the above two questions woud be trivia if the cache size at each BS is arge enough to store the entire fie ibrary in the network, in which case no backhau transmission is needed. This paper considers a more reaistic scenario where the network operator has a fixed budget to depoy ony a imited amount of tota cache size C. Because of the imited cache size, each BS can ony cache a subset of the fies. In the next section, we define the fie deivery performance in the backhau ink in terms of both the deivery rate and the downoading time, which are expressed as functions of cache sizes at the BSs. III. BROADCAST CHANNEL WITH RECEIVER CACHING In this section, we investigate the optima caching strategy for the backhau network with given cache size at each BS. We then formuate the two-stage joint cache and beamforming design probem considered in this paper in the next section. A. Separate Cache-Channe Coding Without BS caching, the downink C-RAN wireess backhau network with a singe custer of BSs as shown in Fig. can be modeed as a broadcast channe BC) with common message ony, whose capacity is given as R 0 Ix;y ), L, 2) where R 0 denotes the muticast rate, Ix;y ) is the mutua information between the transmit signa x at the CP and the received signa y at BS. It can be seen from 2) that the common information rate is imited by the worst channe across the BSs. To dea with the channe disparity issue in 2), this paper considers the use of BS caching to smooth out the difference

4 4 in channe quaity across the BSs. Assuming that BS has cache size C bits, fied up by caching the first C bits of a fie with a tota size of F bits, a simpe caching strategy is to et the CP deiver ony the rest F C bits of the fie to BS. However, since the BSs are served through muticasting, the CP has to send the maximum of the rest of the requested fie, i.e., max {F C }, to make sure that the BS with east cache size can get the entire fie. Assuming that the channe coherent bock is arge enough so that the fie can be competey downoaded within one coherent bock, then the amount of time needed to finish the fie downoading is T 0 = max {F C } min {Ix;y )} and the effective fie downoading rate is 3) D 0 = F T 0 = min {Ix;y )} max { C /F}. 4) As we can see from 3) or 4), with this naive caching strategy, it is optima to aocate the cache size uniformy among the BSs, i.e., C = C/L, L. B. Joint Cache-Channe Coding It is possibe to significanty improve the naive separate cache-channe coding strategy by considering cached content as side information for the broadcast channe. The achievabe rate of this strategy, named as joint cache-channe coding in [23], can be characterized as beow. Lemma [23]): Consider a BC with common message, if receiver L cachesα 0 α ) fraction of the message, then the muticast common message rateris achievabe if and ony if the foowing set of inequaities are satisfied: R α ) Ix;y ), L, 5) where x is the input and y s are the output of the broadcast channe. Proof: We outine an information-theoretic proof as foows. Consider that a message w is chosen uniformy from the index set {,2,...,W} and is to be transmitted to a set of receivers L over n channe uses at a rate of R = ogw n bits per channe use. A codebook C of size [ 2 nr,n ] is first generated by drawing a symbos x i j),i =,2,...,n, and j =,2,...,2 nr, independenty and identicay according to the channe input distribution, where each row of C corresponds to a codeword. To send the message w, the w-th row of C, denoted as X n w) = [x w) x 2 w)... x n w)], is transmitted over the channe. Note that the codebook C is reveaed to both the transmitter and the L receivers. After receiving Y n, the receiver tries to decode the index w by ooking for a codeword in the codebook C that is jointy typica with Y n and the cached content. Suppose that each receiver caches a fraction of the message specificay caches the first α ogw bits of w. Then, receiver ony needs to search among those codewords whose indices start with the same α ogw bits as the cached In a simiar vein, a reated probem formuation of using secondary backhau inks to compensate for channe disparity is investigated in [27]. bits. Since there exist a tota number of 2 nr α ogw such codewords, by the packing emma [28], receiver woud be abe to find the correct codeword with diminishing error probabiity in the imit n as ong as the inequaity nr α ogw nix;y ) is satisfied, or equivaenty R α R Ix;y ), where x is the input channe symbo. This inequaity needs to be satisfied by a L to ensure that the common message is recovered by a the receivers, which eads to the proof of the achievabiity of 5) in Lemma. For the proof of converse, we refer to [23] for the detais. In the setup of this paper, given cache size aocation C and the fie size F, each BS can cache C /F fraction of the fie. Hence, by Lemma, the fie deivery rate D c with the joint cache-channe coding strategy can be formuated as D c = min { Ix;y ) C /F }, 6) and the downoading time T c can be written as T c = F { } F C = max. 7) D c Ix;y ) Ceary, the above fie downoading time and deivery rate are stricty better than the ones in 3) and 4) except when a Ix;y ) are equa to each other. Instead of aocating the cache size C uniformy, 6) and 7) suggest that it is advantageous to aocate more cache to the BSs with weaker channes to achieve an overa higher muticast rate or shorter downoading time. The difficuty, however, ies in the fact that in practice the channe condition changes over time whie the cache size aocations among the BSs can ony be optimized ahead of time at the cache depoyment phase. In the next section, we formuate a two-stage optimization probem that jointy optimizes the cache size aocation strategy based on the ong-term channe statistics and the beamforming strategy based on the short-term channe reaization. IV. TWO-STAGE CACHING AND BEAMFORMING DESIGN We are now ready to formuate the two-stage joint cache size aocation and beamforming design probem. At a sow timescae, cache size aocation is done at the cache depoyment phase, so they can ony adapt to the channe statistics. At a fast timescae, the beamforming vector can be designed to adapt to each channe reaization during the fie deivery phase. First, we fix cache size aocation and content pacement and focus on the beamforming design in the fast timescae. Assuming that the BS uses a singe-datastream muticast beamforming strategy for the mutipe-antenna BC ), the transmit signa is given by x = ws, where w C M is the beamformer vector and s C is the user message, which can be assumed to be compex Gaussian distributedcn 0, ). Then, the mutua information in the previous section becomes Ix;y ) = og + TrH W) ), 8) where H = h h H, and W = ww H is the beamforming covariance matrix of the transmit signa x restricted within the constraint set W = {W 0 TrW) P, rankw) = } 9)

5 5 with P being the transmit power budget at the CP. The above set is nonconvex due to the rank-one constraint. To obtain a numerica soution, a common practice is to drop the rank-one constraint to enabe convex optimization, then to recover a feasibe rank-one beamformer from the resuting soution [29]. Whie the soution so obtained is not necessariy goba optimum, this strategy often works very we in practice, when compared to the gobay optima branch-and-bound agorithm [30]. Under fixed channe reaization H and cache size C, the optima beamformer design probem, after dropping the rankone constraint, can now be formuated in terms of maximizing the deivery rate or equivaenty minimizing the downoading time): maximize {W} D c subject to TrW) P, W 0, 0a) 0b) which can be reformuated as the foowing convex optimization probem: maximize ξ a) {W, ξ} subject to og + TrH ) W) ξf C ), L, b) TrW) P, W 0. c) This probem can be soved efficienty using standard optimization toobox such as CVX [3]. To obtain a rank-one muticast beamforming vector afterwards, we can adopt a strategy of using the eigenvector corresponding to the argest eigenvaue of soution W. The simuation section of this paper ater examines the performance oss due to such a reaxation of the rank-one constraint. Next, we consider the aocation of cache sizes in the sow timescae. The chaenge is now to find the optima aocation C that s the expected fie downoading time or maximizes the expected fie downoading rate over the channe distribution. Intuitivey, the roe of caching at the BSs is to even out the channe capacity disparity in the CP-to-BS inks so as to improve the muticast rate, which is the minimum capacity across the BSs. At the fast timescae, transmit beamforming aready does so to some extent. BS caching aims to further improve the minimum. The chaenge here is to optimize the cache size aocation, which is done in the sow timescae, whie accounting for the effect of beamforming, which is done in the fast timescae as a function of the instantaneous channe. We note that the caching strategy outined in Lemma is universa in the sense that it depends ony on C and not on H. In the next two sections, we devise efficient agorithms that optimize the cache size aocation at the BSs based on the ong-term channe statistics using a sampe approximation technique. V. CACHE ALLOCATION OPTIMIZATION ACROSS THE BSS In this section, we formuate the cache size aocation probem for deivering a singe fie case of fixed size F bits in order to iustrate a sampe approximation technique that aows us to quantify how much cache shoud be aocated to the BSs with different average channe strengths. The mutipe fies case is treated in Section VI. A. Minimizing Expected Downoading Time For given cache size C, the optima fie downoading time 7) can be written as Tc = minmax F C W og + TrH W). 2) ) Note that the fie downoading time has aso been considered in [6], [32] as the objective function. Differenty, in this paper, we take the expectation oftc over the channe distribution and aim to find an optima cache size aocation that s the ong-term expected fie downoading time. The cache optimization probem under a tota cache size constraint C across the BSs is formuated as: {C } subject to E {H }[Tc] C C, 0 C F, L. L 3a) 3b) Finding a cosed-form expression for the objective function in 3a) is difficut. This paper proposes to repace the objective function in 3a) with its sampe approximation [25] and to reformuate the probem as: {C, W n } subject to N N max F C og + TrHn Wn ) ) C C, 0 C F, L, TrW n ) P, W n 0, n N, 4a) 4b) 4c) where N is the sampe size, N := {,2,...,N}, {H n } n N are the channe sampes drawn according to the distribution of H, and W n is the beamforming covariance matrix adapted to the sampes{h n } L. Note that we do not assume any specific channe distribution here. In fact, the above sampe approximation scheme works for any genera channe distribution. Furthermore, even if in practice when the channe distribution is unknown, we can sti use the historica channe reaizations as the channe sampes, as ong as they are samped from the same distribution. Probem 4) is sti not easy to sove mainy due to the foowing two reasons. First, the objective function of probem 4) is nonsmooth and nonconvex, abeit a of its constraints are convex. Second, the sampe size N generay needs to be sufficienty arge such that the sampe average is a good approximation to the origina expected downoading time, eading to a high compexity for soving probem 4) directy. In the foowing, we first reformuate probem 4) as a smooth probem and inearize the nonconvex term, then everage the ADMM approach to decoupe the probem inton

6 6 ow-compexity convex subprobems to improve the efficiency of soving the probem. First, drop the constant /N in 4a) and introduce the auxiiary variabe {ξ n }, and reformuate probem 4) as {C, W n, ξ n } subject to N ξ n og + TrHn Wn ) 4b) and 4c). ) ξ n F C ), 5a) L, n N, 5b) The above probem 5) is smooth but sti nonconvex due to constraint 5b). To dea with this nonconvex constraint, we approximate the nonconvex term ξ n F C ) in 5b) by its first-order Tayor expansion at some appropriate point ξ n, C ), i.e., ξ n F C ) ξ n F C )+ [ F C, ξ n] [ ξ n ξ n, C C ] T 6) = ξ n F C )+ F C ) ξ n ξ n). 7) Based on 7), an iterative first-order approximation is proposed in Agorithm for soving probem 4). More specificay, et {ξ n t), C t)} be the iterates at the t-th iteration, the agorithm soves {C, W n, ξ n } N subject to og ξ n + TrHn Wn ) ) ξ n t)f C ) 8a) +F C t))ξ n ξ n t)), L, n N, 8b) ξ n ξ n t) rt), n N, 8c) C C t) rt), L, 8d) 4b) and 4c), with fixed {ξ n t),c t)}, where 8c) and 8d) are the trust region constraints [33], within which we trust that the inear approximation in 8b) is accurate, and rt) is the trust region radius at the t-th iteration, which is chosen in a way such that the foowing condition is satisfied: N N ξ n t) N max F C t) og + TrHn Wn t)) N ξ n t) ξ n t) ) τ, 9) where W n t), C t), ξn t) are soutions to probem 8) and τ 0,) is a constant. Notice that the numerator in 9) is the actua reduction in the objective of probem 4) and the denominator is the predicted reduction. The condition in 9) basicay says that the trust region radius is accepted ony if the ratio of the actua reduction and the predicted reduction Agorithm Optimized Cache Aocation with Singe Fie Initiaization: Initiaize C ) = C/L, L, and ξ n ) as the soution to probem 5) with C = C ); set t = ; Repeat: ) Initiaize the trust region radius rt) = ; Repeat: a) Use the ADMM approach in Appendix B to sove probem 8); b) Update rt) = rt)/2; Unti condition 9) is satisfied. 2) Update {ξ n t+),c t+)} according to 20) and 2), respectivey; 3) Set t = t+; Unti convergence is greater than or equa to a constant, in which case probem 8) is a good approximation of the origina probem 4). After soving probem 8), the agorithm updates the parameters for the next iteration by substituting the soution obtained from the ineary approximated probem 8) to the origina probem 4): ξ n t+) = min L og ) + TrHn Wn t)) F C t), n N, 20) C t+) = C t), L. 2) For the initia point, we can set C ) to be C/L for a L, then probem 5) can be decouped into N convex optimization subprobems to sove for ξ n ) for a n N. It remains to sove probem 8). Note that probem 8) is a convex probem but with a potentiay arge number of variabes due to the arge sampe size. We propose an ADMM approach [26] to sove probem 8), which decoupes the high-dimensiona probem into N decouped smadimensiona subprobems. The detais of soving probem 8) using the ADMM approach can be found in Appendix B. It can be shown that the ADMM approach is guaranteed to converge to the goba optimum soution of the convex optimization probem 8). Once probem 3) is soved using Agorithm, we fix the obtained optimized cache size aocation and evauate its effectiveness under a different set of independenty generated channes and cacuate the fie downoading time 2 for each channe by soving the convex probem ). Agorithm is guaranteed to converge to a stationary point of the optimization probem 4). In the rest of this section, we prove the convergence of Agorithm. First, we define the stationary point of probem 4) as in [34]. Definition : Consider a more genera probem x X Fx) 22) 2 Note that the optima downoading time for each given channe is the inverse of the optima objective vaue of probem ).

7 7 where X is the feasibe set and Fx) is defined as Fx) := N N max{f n x)}. 23) Here, {f n x)} is a set of continuousy differentiabe functions. Given any feasibe point x, define { Φ x) = max { d, x + d X} N N F x) { max fn x)+ f n x) T d }}. 24) A point x X is caed a stationary point of probem 22) if Φ x) = 0. Two remarks on the above definition of the stationary point are in order. First, it is simpe to see that Φ x) is aways nonnegative as d = 0 is a feasibe point of 24). If Φ x) = 0, it means that there does not exist any feasibe and decreasing direction at point x in the first-order approximation sense. Second, probem 4) is in the form of probem 22) if we set ) x = {C,W n }, f n x) = F C )/og + TrHn Wn ) σ, 2 and X to be the feasibe set of probem 4), which is convex and bounded. Based on the above stationary point definition, we now state the convergence resut of Agorithm in the foowing theorem. Theorem : Agorithm is guaranteed to converge. Any accumuation point of the sequence generated by Agorithm is a stationary point of probem 4), or equivaenty probem 5). Proof: Agorithm is a specia case of the genera nonsmooth trust region agorithm discussed in [35, Chapter ], which can be proved to converge to a stationary point of the genera probem 22). For competeness of this paper, we provide a proof outine in Appendix A. B. Maximizing Expected Downoading Rate In this subsection, we consider maximizing the expected fie downoading rate as the objective function to optimize the BS cache size aocation, which can be formuated as [ ] F maximize E {H } {C } Tc 25a) subject to C C, 0 C F, L, 25b) L where Tc is the optima fie downoading time defined in 2) under given channe reaization and cache size aocation. Note that the expected vaue of the inverse of a random variabe X, E [ X], is in genera different from the inverse of the expected vaue ofx, E[X]. Thus, the cache size aocation obtained from soving probem 25) is aso different from the one obtained from soving probem 3). We use the same idea as in the previous subsection to sove probem 25). First, we repace the objective function 25a) with its sampe approximation and reformuate the probem as N maximize {C, W n, ξ n } ξ n 26a) subject to 4b), 4c), and 5b), in which we have dropped the constants N and F from the objective function. Then, we repace the nonconvex term in constraint 5b) by its inear approximation 7) and sove probem 26) via optimizing a sequence of ineary approximated probems simiar to probem 8). The approximated probem at each iteration is soved via an ADMM approach simiar to the one described in Appendix B with the ony difference being that the first term ξ in the subprobem 37) n needs to be repaced by ξ n. Same as in the previous subsection, once the optimized cache size aocation is obtained from soving probem 26), we evauate its effectiveness on different sets of channes and sove the muticast rate 3 by optimizing probem ). VI. CACHE ALLOCATION OPTIMIZATION ACROSS FILES In this section, we consider the cache size aocation probem for the genera case with mutipe fies having different popuarities. Due to the minima difference between the downoading rate and the downoading time as described in the previous section, we ony focus on minimizing the expected fie downoading time as the objective function in this section. We assume that each fie k of equa size F bits is requested from the user with given probabiity p k,k K := k p k =, and that BS caches C k /F {,2,...,K}, fraction of fie k with a tota cache size constraint given by,k C k C. Given that fie k is requested, according to Lemma, the optima downoading time for fie k, denoted as Tk, can be written as Tk = min max W k W F C k og + TrH W k ) ). 27) Different from the downoading time 2) in the singe fie case, the above optima downoading time Tk is a random variabe depending on not ony the channe reaization but aso the index of the requested fie. We take the expected vaue of Tk on both the channe reaization H and the fie index k as the objective function and formuate the muti-fie cache size aocation probem as p k E {H }[Tk] 28a) {C k } k subject to C k C, 0 C k F, L, k K.,k 28b) Athough intuitivey the more popuar fie shoud be aocated arger cache size, the question of how much cache shoud be aocated to each fie is nontrivia. In particuar, it is in 3 The optima muticast rate for given channe is the optima objective vaue of probem ).

8 8 genera not true that one shoud aocate the most popuar fie in its entirety first, then the second most popuar fie, etc., unti the cache size is exhausted. This is because the gain in term of the objective function of the optimization probem 28) due to aocating progressivey more cache size to one fie diminishes as more cache is aocated. At some point, it is better to aocate some cache to the ess popuar fies, even when the most popuar fie has not been entirey cached. The optima aocation needs to be found by soving probem 28). To sove probem 28), we use the same sampe approximation idea as in the singe fie case. With an additiona set of axiary variabes ξk n, probem 28) after the sampe approximation can be formuated as {C k, W n k, ξn k} subject to K N p k k= og ξ n k + TrHn k Wn k ) 29a) ) ξk n F C k), L, n N, k K, 29b) C k C, 0 C k F,,k TrW n k ) P, Wn k 0, L, k K, 29c) n N, k K. 29d) Probem 29) is then soved in an iterative fashion. At each iteration the nonconvex term on the right hand side of 29b) is repaced by its first-order approximation and the resuting convex probem to be soved at the t-th iteration is given by: {C k, W n k, ξn k} K N p k k= subject to og ξ n k + TrHn k Wn k ) 30a) ) ξk n t)f C k) +F C k t))ξ n k ξ n kt)), L, n N, k K, 30b) ξ n k ξ n kt) rt), n N, k K, 30c) C k C k t) rt), L, k K, 30d) 29c) and 29d), where ξ n k t) and C kt) are fixed parameters obtained from the previous iteration and are updated for the next iteration according to ξ n kt+) = min L og ) + TrHn k Wn k t)) F C k t), n N, k K, 3) C k t+) = C kt), L, k K, 32) where Wk n t) and C k t) are soutions to probem 30). Simiar to 9), the trust region radius rt) in 30c) and 30d) Agorithm 2 Optimized Cache Aocation with Mutipe Fies Initiaization: Initiaize C k ) = C/LK, L, k K and ξ n k ) as the soution to probem 29) with C k = C k ); set t = ; Repeat: ) Initiaize the trust region radius rt) = ; Repeat: a) Use the ADMM approach in Appendix C to sove probem 30); b) Update rt) = rt)/2; Unti condition 33) is satisfied. 2) Update {ξ n k t+),c kt+)} according to 3) and 32), respectivey; 3) Set t = t+; Unti convergence is picked to satisfy the foowing condition: K N p k F C ξ n max k t) k= k t) og k= + TrHn k Wn t)) ) k K N p k ξk nt) ) ξk n t) τ 33) for some constant τ 0,). Note that probem 30) can aso be soved by using an ADMM approach as expained in Appendix C, which decoupes probem 30) into NK subprobems and each subprobem corresponds to a pair of sampe channe and fie request. The overa proposed agorithm for soving the cache size aocation probem with mutipe fies is summarized in Agorithm 2. Once the cache size aocation for the mutipe fies case is optimized through Agorithm 2, we cacuate the downoading time for fie k by soving probem 27) with fixed C k, which can be formuated as a convex optimization probem simiar to ). We then compute the average downoading time by averaging under different sets of channe reaizations. VII. SIMULATION RESULTS This section evauates the performance of our proposed caching schemes through simuations. Consider a downink C-RAN mode with L = 5 BSs randomy paced on the haf pane beow the CP with the reative distances between the CP and the 5 BSs shown in Fig. 2. We generate 000 sets of channe reaizations from the CP to the BSs according to h = K /2 v, where K modes the correation between the CP transmit antennas to BS and is generated mainy according to the ange-of-arriva and the antenna pattern, as described in [36], with the path-oss component modeed as og 0 d) db and d is the distance between the coud and the BS in kiometers;v is a Gaussian random vector with each eement independenty and identicay distributed as CN 0,). The first N = 00 sets of channe reaizations are

9 Coud Uniform Proportiona Optimized - Downoading Time Optimized - Downoading Rate -00 meter BS 5 BS 4 BS 3 BS 2 BS Cache Size meter 0 BS BS 2 BS 3 BS 4 BS 5 Fig. 2. A downink C-RAN setup with 5 BSs. The distances from the CP to the 5 BSs are 398, 278, 473, 286, 267) meters, respectivey. Fig. 3. Cache aocation for different schemes under tota cache size C = 00, normaized with respect to fie size F = 00. TABLE I SIMULATION PARAMETERS. Parameters Vaues Number of BSs 5 Backhau channe bandwidth 20 MHz Number of antennas at CP 0 Number of antennas at each BS Maximum transmit power P at CP 40 Watts Antenna gain 7 dbi Background noise 50 dbm/hz Path oss from CP to BS og 0 d) Rayeigh sma scae fading 0 db Normaized fie size 00 Training sampe size N 00 Test sampe size 900 used in the sampe approximation to optimize the cache aocation whie the rest 900 are used to evauate the performance under the obtained cache size aocation. The detais of the simuation parameters are isted in Tabe I. A. Cache Aocation for BSs with Varying Channe Strengths In this subsection, we evauate the performances of the proposed schemes for caching a singe fie across mutipe BSs with different channe strengths as discussed in Section V. We compare the optimized cache size aocations obtained from minimizing the expected fie downoading time 3) and maximizing the expected fie downoading rate 25) with the foowing set of schemes: No Cache: Cache sizes C = 0 for a BSs; Uniform Cache Aocation: Cache sizes among the BSs are uniformy distributed as C = C/L, which serves as a baseine scheme; Proportiona Cache Aocation: Cache sizes among the BSs are proportionay ) aocated such that F C )/og + P TrK ) Lσ for a are equaized, if 2 possibe, which serves as another baseine scheme; Lower/Upper Bound: Cache sizes among the BSs are dynamicay and optimay aocated by soving probem ) for each channe reaization by treating {C } as the optimization variabes, which is impractica in reaity but serves as a ower bound for minimizing the expected fie downoading time and an upper bound for maximizing the expected fie downoading rate; Rank-One Muticast Beamformer: Cache sizes among the BSs are the same as the optimized caching schemes, but the muticast beamformer is restricted to be rank-one and is set to be the eigenvector corresponding to the argest eigenvaue of the optimized beamforming matrix W n in each test sampe channe. In Fig. 3, we compare the aocated BS cache sizes between the proposed schemes trained on the first 00 channes and the baseine schemes under normaized fie size F = 00 and tota cache size C = 00. As we can see, both of the proposed caching schemes are more aggressive in aocating arger cache sizes to the weaker BS 3 as compared to the uniform and proportiona caching schemes. We then evauate the performances of different cache size aocation schemes on the rest 900 sampe channes and report the fie downoading time and downoading rate or spectra efficiency) in Tabe II and III, respectivey, under two different settings of tota cache size C = 00 and C = 200, normaized with respect to fie size F = 00. As we can see, the proposed caching scheme improves over the uniform and proportiona caching schemes by 0% 5% on average, but the gains are more significant for the 90th-percentie downoading time and the 0th-percentie downoading rate, which are around 20% 27% and 26% 36%, respectivey. We note here that without caching, the average and 90thpercentie fie downoading time are.45 ms/mb and 4.76 ms/mb, respectivey, in this setting. The average and 0thpercentie fie downoading rate are 4.63 bps/hz and 3.39 bps/hz. Thus, the optimized BS caching schemes with C = 00 and C = 200 normaized with respect to F = 00) improve the average downoading time by about 33% and 50% respectivey, and improve the average downoading rate

10 0 TABLE II FILE DOWNLOADING TIME MS/MB) COMPARISON FOR DIFFERENT TOTAL CACHE SIZES, NORMALIZED WITH RESPECT TO FILE SIZE F = 00. Cache Scheme Tota Cache C = 00 Tota Cache C = 200 Average 90th-Percentie Average 90th-Percentie Uniform Proportiona Optimized Rank-One Lower Bound TABLE III FILE DOWNLOADING RATE BPS/HZ) COMPARISON FOR DIFFERENT TOTAL CACHE SIZES, NORMALIZED WITH RESPECT TO FILE SIZE F = 00. Cache Scheme Tota Cache C = 00 Tota Cache C = 200 Average 0th-Percentie Average 0th-Percentie Uniform Proportiona Optimized Rank-One Upper Bound Cumuative Distribution Function No Cache, C = 0 Uniform, C = 00 Proportiona, C = 00 Optimized, C = 00 Lower Bound, C = 00 Rank-One, C = 00 Uniform, C = 200 Proportiona, C = 200 Optimized, C = 200 Lower Bound, C = 200 Rank-One, C = 200 Cumuative Distribution Function No Cache, C = 0 Uniform, C = 00 Proportiona, C = 00 Optimized, C = 00 Upper Bound, C = 00 Rank-One, C = 00 Uniform, C = 200 Proportiona, C = 200 Optimized, C = 200 Upper Bound, C = 200 Rank-One, C = Downoading Time in ms/mb Fig. 4. CDF of downoading time under different caching schemes with tota cache size C = 00 and C = 200, respectivey, normaized with respect to fie size F = Muticast Rate in bps/hz Fig. 5. CDF of downoading rates under different caching schemes with tota cache size C = 00 and C = 200, respectivey, normaized with respect to fie size F = 00. by about 43% and 9% respectivey. In Figs. 4 and 5, we compare the cumuative distribution functions CDFs) of the downoading time and the downoading rates evauated on the 900 test channes with different caching schemes. Simiar to what we have seen in Tabes II and III, the proposed caching scheme shows significant gain on the high downoading time regime in Fig. 4 and on the ow downoading rate regime in Fig. 5 as compared to the baseine schemes. From Figs. 4 and 5, we can aso see that the rankone muticast beamformer shows negigibe performance oss as compared to the genera-rank muticast beamformer matrix W n obtained by soving ). It is aso worth remarking that the ower bound scheme in Fig. 4 and the upper bound scheme in Fig. 5 sove the cache size aocation probem dynamicay for each channe reaization, which is impractica, and ony serve as benchmark schemes in this paper. To summarize the insight from the simuation resuts in this subsection for the singe fie case: First, athough both the uniform and the proportiona caching schemes perform fairy we in terms of the average fie downoading time and downoading rate, the proposed caching scheme shows significant gains in improving the high downoading time regime and the ow downoading rate regime. This is due to the fact that BSs farther away from the coud are more aggressivey aocated arger amount of cache under the optimized scheme. Second, the rank-one beamformer derived from the generarank covariance matrix does not degrade the performance much at a. Hence, we ony focus on the performance of the proposed caching schemes without the rank- constraint on the covariance matrix in the next subsection for the mutipe fies case. B. Cache Aocation for Fies of Varying Popuarities In this subsection, we present simuation resuts for the caching schemes with mutipe fies having different popuarities and focus on the expected fie downoading time as the performance metric. We first consider ony two fies with different pairs of request probabiities p,p 2 ) isted on

11 TABLE IV OPTIMIZED CACHE ALLOCATIONC,C 2 ) FOR A 2-FILE CASE WITH DIFFERENT FILE POPULARITIES UNDERC = 00 AND F = 00. Fie Popuarity p,p 2 ) = 0.5,0.5) p,p 2 ) = 0.6,0.4) p,p 2 ) = 0.7,0.3) p,p 2 ) = 0.8,0.2) p,p 2 ) = 0.9,0.) BS 8.2, 8.2) 3.2, 2.7) 6.8, 0) 20.2, 0) 22.2, 0) BS2 0, 0) 0, 0) 0, 0) 4.6, 0) 7., 0) BS3 4.8, 4.8) 48.2, 35.9) 53.6, 27) 56.8, 0.9) 58.8, 0) BS4 0, 0) 0, 0) 2.6, 0) 7.5, 0) 0., 0) BS5 0, 0) 0, 0) 0, 0) 0, 0).8, 0) Tota 50, 50) 6.4, 38.6) 73, 27) 89., 0.9) 00, 0) the first row of Tabe IV, where each coumn denotes the cache size aocation among the 5 BSs under the specific fie popuarity given in the first row and each ce gives the cache size aocation between the two fies within each BS. The cache sizes in each coumn add up to the tota cache size C = 00, normaized with respect to fie size F = 00. From Tabe IV we see that for each coumn with given fie popuarity, the weakest BS 3 aways gets the most cache size as in the singe fie case shown in Fig. 3. Moreover, as the difference between the popuarities of the two fies increases across the coumns, more cache is aocated to the first fie. For exampe, the proposed caching scheme decides to aocate a the cache to ony the more popuar fie when p,p 2 ) = 0.9,0.). In Fig. 6, we compare the average fie downoading time between the optimized cache scheme and the foowing baseine schemes: No Cache: Cache size C k = 0 for a BSs and fies; Uniform Cache Aocation: Cache size for fie k at each BS is set to be as C k = C/LK for a k and ; Proportiona Cache Aocation: We first set the tota cache size aocated for fie k as p k C, then distribute p k C among the BSs according to the rue descried in the Proportiona Cache Aocation scheme in Section VII-A; Caching the Most Popuar Fie: We cache the most popuar fie in its entirety first, then the second most popuar fie, etc. When a fie cannot be cache entirey, we distribute the remaining cache among the BSs according to the Proportiona Cache Aocation scheme described in Section VII-A. In Fig. 6, we fix the number of fies to be K = 4 and generate the fie popuarity according to the Zipf distribution [37] given by p k = k α K, k, with different settings of α. As the i= Zipf distribution i α exponent α increases, the difference among the fie popuarities aso increases. As we can see from Fig. 6, the average downoading time for a schemes, except for the uniform caching scheme, decreases as α increases. This is because in uniform cache aocation the cache size is the same for a fies, hence the downoading time is the same no matter which fie is requested. In contrast, a other three schemes tend to aocate more cache to the more popuar fies. In particuar, the proposed caching scheme converges to the scheme of caching the most popuar fie when α =.5, whie it consistenty outperforms the proportiona caching scheme. From Fig. 6 we concude that first, the uniform cache size aocation scheme performs poory when the fies have different popuarities and especiay when the difference is Average Downoading Time in ms/mb No Cache Uniform Proportiona Most Popuar Optimized Zipf Distribution Exponent α Fig. 6. Average downoading time for different Zipf fie distributions under the same number of fies K = 4 and tota cache size C = 400, normaized with respect to fie size F = 00. arge. Second, it is advantageous to aocate arger cache size to the more popuar fie, however, it is not trivia to decide how much more cache is needed for the more popuar fie. Our proposed caching scheme provides a better cache size aocation soution as compared to the heuristic proportiona caching scheme and the most popuar fie caching scheme. VIII. CONCLUSION This paper points out that caching can be used to even out the channe disparity in a muticast scenario. We study the optima BS cache size aocation probem in the downink C- RAN with wireess backhau to iustrate the advantage of muticast and caching for the data-sharing strategy. We first derive the optima muticast rate with BS caching, then formuate the cache size optimization probem under two objective functions, minimizing the expected fie downoading time and maximizing the expected fie downoading rate, subject to the tota cache size constraint. By everaging the sampe approximation method and ADMM, we propose efficient cache size aocation agorithms that consideraby outperform the heuristic schemes. APPENDIX A PROOF OF THEOREM We use the notations introduced in Definition in the foowing convergence proof. First of a, it is simpe to show that the objective sequence {Fxt))} generated by Agorithm monotonicay decreases and is ower bounded by zero.

CLOUD RADIO ACCESS NETWORK WITH OPTIMIZED BASE-STATION CACHING

CLOUD RADIO ACCESS NETWORK WITH OPTIMIZED BASE-STATION CACHING CLOUD RADIO ACCESS NETWORK WITH OPTIMIZED BASE-STATION CACHING Binbin Dai and Wei Yu Ya-Feng Liu Department of Eectrica and Computer Engineering University of Toronto, Toronto ON, Canada M5S 3G4 Emais:

More information

A Design Method for Optimal Truss Structures with Certain Redundancy Based on Combinatorial Rigidity Theory

A Design Method for Optimal Truss Structures with Certain Redundancy Based on Combinatorial Rigidity Theory 0 th Word Congress on Structura and Mutidiscipinary Optimization May 9 -, 03, Orando, Forida, USA A Design Method for Optima Truss Structures with Certain Redundancy Based on Combinatoria Rigidity Theory

More information

Mobile App Recommendation: Maximize the Total App Downloads

Mobile App Recommendation: Maximize the Total App Downloads Mobie App Recommendation: Maximize the Tota App Downoads Zhuohua Chen Schoo of Economics and Management Tsinghua University chenzhh3.12@sem.tsinghua.edu.cn Yinghui (Catherine) Yang Graduate Schoo of Management

More information

Neural Network Enhancement of the Los Alamos Force Deployment Estimator

Neural Network Enhancement of the Los Alamos Force Deployment Estimator Missouri University of Science and Technoogy Schoars' Mine Eectrica and Computer Engineering Facuty Research & Creative Works Eectrica and Computer Engineering 1-1-1994 Neura Network Enhancement of the

More information

Nearest Neighbor Learning

Nearest Neighbor Learning Nearest Neighbor Learning Cassify based on oca simiarity Ranges from simpe nearest neighbor to case-based and anaogica reasoning Use oca information near the current query instance to decide the cassification

More information

Quality of Service Evaluations of Multicast Streaming Protocols *

Quality of Service Evaluations of Multicast Streaming Protocols * Quaity of Service Evauations of Muticast Streaming Protocos Haonan Tan Derek L. Eager Mary. Vernon Hongfei Guo omputer Sciences Department University of Wisconsin-Madison, USA {haonan, vernon, guo}@cs.wisc.edu

More information

A Petrel Plugin for Surface Modeling

A Petrel Plugin for Surface Modeling A Petre Pugin for Surface Modeing R. M. Hassanpour, S. H. Derakhshan and C. V. Deutsch Structure and thickness uncertainty are important components of any uncertainty study. The exact ocations of the geoogica

More information

Interference Spins Popovski, Petar; Simeone, Osvaldo; Nielsen, Jimmy Jessen; Stefanovic, Cedomir

Interference Spins Popovski, Petar; Simeone, Osvaldo; Nielsen, Jimmy Jessen; Stefanovic, Cedomir Aaborg Universitet Interference Spins Popovski, Petar; Simeone, Osvado; Niesen, Jimmy Jessen; Stefanovic, Cedomir Pubished in: I E E E Communications Letters DOI (ink to pubication from Pubisher): 10.1109/LCOMM.2014.2387166

More information

Language Identification for Texts Written in Transliteration

Language Identification for Texts Written in Transliteration Language Identification for Texts Written in Transiteration Andrey Chepovskiy, Sergey Gusev, Margarita Kurbatova Higher Schoo of Economics, Data Anaysis and Artificia Inteigence Department, Pokrovskiy

More information

Alternative Decompositions for Distributed Maximization of Network Utility: Framework and Applications

Alternative Decompositions for Distributed Maximization of Network Utility: Framework and Applications Aternative Decompositions for Distributed Maximization of Network Utiity: Framework and Appications Danie P. Paomar and Mung Chiang Eectrica Engineering Department, Princeton University, NJ 08544, USA

More information

Lecture outline Graphics and Interaction Scan Converting Polygons and Lines. Inside or outside a polygon? Scan conversion.

Lecture outline Graphics and Interaction Scan Converting Polygons and Lines. Inside or outside a polygon? Scan conversion. Lecture outine 433-324 Graphics and Interaction Scan Converting Poygons and Lines Department of Computer Science and Software Engineering The Introduction Scan conversion Scan-ine agorithm Edge coherence

More information

On Upper Bounds for Assortment Optimization under the Mixture of Multinomial Logit Models

On Upper Bounds for Assortment Optimization under the Mixture of Multinomial Logit Models On Upper Bounds for Assortment Optimization under the Mixture of Mutinomia Logit Modes Sumit Kunnumka September 30, 2014 Abstract The assortment optimization probem under the mixture of mutinomia ogit

More information

Solving Large Double Digestion Problems for DNA Restriction Mapping by Using Branch-and-Bound Integer Linear Programming

Solving Large Double Digestion Problems for DNA Restriction Mapping by Using Branch-and-Bound Integer Linear Programming The First Internationa Symposium on Optimization and Systems Bioogy (OSB 07) Beijing, China, August 8 10, 2007 Copyright 2007 ORSC & APORC pp. 267 279 Soving Large Doube Digestion Probems for DNA Restriction

More information

Chapter Multidimensional Direct Search Method

Chapter Multidimensional Direct Search Method Chapter 09.03 Mutidimensiona Direct Search Method After reading this chapter, you shoud be abe to:. Understand the fundamentas of the mutidimensiona direct search methods. Understand how the coordinate

More information

Topology-aware Key Management Schemes for Wireless Multicast

Topology-aware Key Management Schemes for Wireless Multicast Topoogy-aware Key Management Schemes for Wireess Muticast Yan Sun, Wade Trappe,andK.J.RayLiu Department of Eectrica and Computer Engineering, University of Maryand, Coege Park Emai: ysun, kjriu@gue.umd.edu

More information

Design of IP Networks with End-to. to- End Performance Guarantees

Design of IP Networks with End-to. to- End Performance Guarantees Design of IP Networks with End-to to- End Performance Guarantees Irena Atov and Richard J. Harris* ( Swinburne University of Technoogy & *Massey University) Presentation Outine Introduction Mutiservice

More information

Extended Node-Arc Formulation for the K-Edge-Disjoint Hop-Constrained Network Design Problem

Extended Node-Arc Formulation for the K-Edge-Disjoint Hop-Constrained Network Design Problem Extended Node-Arc Formuation for the K-Edge-Disjoint Hop-Constrained Network Design Probem Quentin Botton Université cathoique de Louvain, Louvain Schoo of Management, (Begique) botton@poms.uc.ac.be Bernard

More information

Minimizing Resource Cost for Camera Stream Scheduling in Video Data Center

Minimizing Resource Cost for Camera Stream Scheduling in Video Data Center Gao YH, Ma HD, Liu W. Minimizing resource cost for camera stream scheduing in video data center. JOURNAL OF COMPUTER SCIENCE AND TECHNOLOGY 32(3): 555 570 May 2017. DOI 10.1007/s11390-017-1743-x Minimizing

More information

Load Balancing by MPLS in Differentiated Services Networks

Load Balancing by MPLS in Differentiated Services Networks Load Baancing by MPLS in Differentiated Services Networks Riikka Susitaiva, Jorma Virtamo, and Samui Aato Networking Laboratory, Hesinki University of Technoogy P.O.Box 3000, FIN-02015 HUT, Finand {riikka.susitaiva,

More information

A Memory Grouping Method for Sharing Memory BIST Logic

A Memory Grouping Method for Sharing Memory BIST Logic A Memory Grouping Method for Sharing Memory BIST Logic Masahide Miyazai, Tomoazu Yoneda, and Hideo Fuiwara Graduate Schoo of Information Science, Nara Institute of Science and Technoogy (NAIST), 8916-5

More information

Hiding secrete data in compressed images using histogram analysis

Hiding secrete data in compressed images using histogram analysis University of Woongong Research Onine University of Woongong in Dubai - Papers University of Woongong in Dubai 2 iding secrete data in compressed images using histogram anaysis Farhad Keissarian University

More information

For Review Only. CFP: Cooperative Fast Protection. Bin Wu, Pin-Han Ho, Kwan L. Yeung, János Tapolcai and Hussein T. Mouftah

For Review Only. CFP: Cooperative Fast Protection. Bin Wu, Pin-Han Ho, Kwan L. Yeung, János Tapolcai and Hussein T. Mouftah Journa of Lightwave Technoogy Page of CFP: Cooperative Fast Protection Bin Wu, Pin-Han Ho, Kwan L. Yeung, János Tapocai and Hussein T. Mouftah Abstract We introduce a nove protection scheme, caed Cooperative

More information

Further Optimization of the Decoding Method for Shortened Binary Cyclic Fire Code

Further Optimization of the Decoding Method for Shortened Binary Cyclic Fire Code Further Optimization of the Decoding Method for Shortened Binary Cycic Fire Code Ch. Nanda Kishore Heosoft (India) Private Limited 8-2-703, Road No-12 Banjara His, Hyderabad, INDIA Phone: +91-040-3378222

More information

A Fast Block Matching Algorithm Based on the Winner-Update Strategy

A Fast Block Matching Algorithm Based on the Winner-Update Strategy In Proceedings of the Fourth Asian Conference on Computer Vision, Taipei, Taiwan, Jan. 000, Voume, pages 977 98 A Fast Bock Matching Agorithm Based on the Winner-Update Strategy Yong-Sheng Chenyz Yi-Ping

More information

Replication of Virtual Network Functions: Optimizing Link Utilization and Resource Costs

Replication of Virtual Network Functions: Optimizing Link Utilization and Resource Costs Repication of Virtua Network Functions: Optimizing Link Utiization and Resource Costs Francisco Carpio, Wogang Bziuk and Admea Jukan Technische Universität Braunschweig, Germany Emai:{f.carpio, w.bziuk,

More information

Outline. Parallel Numerical Algorithms. Forward Substitution. Triangular Matrices. Solving Triangular Systems. Back Substitution. Parallel Algorithm

Outline. Parallel Numerical Algorithms. Forward Substitution. Triangular Matrices. Solving Triangular Systems. Back Substitution. Parallel Algorithm Outine Parae Numerica Agorithms Chapter 8 Prof. Michae T. Heath Department of Computer Science University of Iinois at Urbana-Champaign CS 554 / CSE 512 1 2 3 4 Trianguar Matrices Michae T. Heath Parae

More information

Lecture Notes for Chapter 4 Part III. Introduction to Data Mining

Lecture Notes for Chapter 4 Part III. Introduction to Data Mining Data Mining Cassification: Basic Concepts, Decision Trees, and Mode Evauation Lecture Notes for Chapter 4 Part III Introduction to Data Mining by Tan, Steinbach, Kumar Adapted by Qiang Yang (2010) Tan,Steinbach,

More information

A METHOD FOR GRIDLESS ROUTING OF PRINTED CIRCUIT BOARDS. A. C. Finch, K. J. Mackenzie, G. J. Balsdon, G. Symonds

A METHOD FOR GRIDLESS ROUTING OF PRINTED CIRCUIT BOARDS. A. C. Finch, K. J. Mackenzie, G. J. Balsdon, G. Symonds A METHOD FOR GRIDLESS ROUTING OF PRINTED CIRCUIT BOARDS A C Finch K J Mackenzie G J Basdon G Symonds Raca-Redac Ltd Newtown Tewkesbury Gos Engand ABSTRACT The introduction of fine-ine technoogies to printed

More information

Modelling and Performance Evaluation of Router Transparent Web cache Mode

Modelling and Performance Evaluation of Router Transparent Web cache Mode Emad Hassan A-Hemiary IJCSET Juy 2012 Vo 2, Issue 7,1316-1320 Modeing and Performance Evauation of Transparent cache Mode Emad Hassan A-Hemiary Network Engineering Department, Coege of Information Engineering,

More information

An improved distributed version of Han s method for distributed MPC of canal systems

An improved distributed version of Han s method for distributed MPC of canal systems Deft University of Technoogy Deft Center for Systems and Contro Technica report 10-013 An improved distributed version of Han s method for distributed MPC of cana systems M.D. Doan, T. Keviczky, and B.

More information

A Novel Congestion Control Scheme for Elastic Flows in Network-on-Chip Based on Sum-Rate Optimization

A Novel Congestion Control Scheme for Elastic Flows in Network-on-Chip Based on Sum-Rate Optimization A Nove Congestion Contro Scheme for Eastic Fows in Network-on-Chip Based on Sum-Rate Optimization Mohammad S. Taebi 1, Fahimeh Jafari 1,3, Ahmad Khonsari 2,1, and Mohammad H. Yaghmae 3 1 IPM, Schoo of

More information

Response Surface Model Updating for Nonlinear Structures

Response Surface Model Updating for Nonlinear Structures Response Surface Mode Updating for Noninear Structures Gonaz Shahidi a, Shamim Pakzad b a PhD Student, Department of Civi and Environmenta Engineering, Lehigh University, ATLSS Engineering Research Center,

More information

An Adaptive Two-Copy Delayed SR-ARQ for Satellite Channels with Shadowing

An Adaptive Two-Copy Delayed SR-ARQ for Satellite Channels with Shadowing An Adaptive Two-Copy Deayed SR-ARQ for Sateite Channes with Shadowing Jing Zhu, Sumit Roy zhuj@ee.washington.edu Department of Eectrica Engineering, University of Washington Abstract- The paper focuses

More information

ACTIVE LEARNING ON WEIGHTED GRAPHS USING ADAPTIVE AND NON-ADAPTIVE APPROACHES. Eyal En Gad, Akshay Gadde, A. Salman Avestimehr and Antonio Ortega

ACTIVE LEARNING ON WEIGHTED GRAPHS USING ADAPTIVE AND NON-ADAPTIVE APPROACHES. Eyal En Gad, Akshay Gadde, A. Salman Avestimehr and Antonio Ortega ACTIVE LEARNING ON WEIGHTED GRAPHS USING ADAPTIVE AND NON-ADAPTIVE APPROACHES Eya En Gad, Akshay Gadde, A. Saman Avestimehr and Antonio Ortega Department of Eectrica Engineering University of Southern

More information

MULTIGRID REDUCTION IN TIME FOR NONLINEAR PARABOLIC PROBLEMS: A CASE STUDY

MULTIGRID REDUCTION IN TIME FOR NONLINEAR PARABOLIC PROBLEMS: A CASE STUDY MULTIGRID REDUCTION IN TIME FOR NONLINEAR PARABOLIC PROBLEMS: A CASE STUDY R.D. FALGOUT, T.A. MANTEUFFEL, B. O NEILL, AND J.B. SCHRODER Abstract. The need for paraeism in the time dimension is being driven

More information

AN EVOLUTIONARY APPROACH TO OPTIMIZATION OF A LAYOUT CHART

AN EVOLUTIONARY APPROACH TO OPTIMIZATION OF A LAYOUT CHART 13 AN EVOLUTIONARY APPROACH TO OPTIMIZATION OF A LAYOUT CHART Eva Vona University of Ostrava, 30th dubna st. 22, Ostrava, Czech Repubic e-mai: Eva.Vona@osu.cz Abstract: This artice presents the use of

More information

5940 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 13, NO. 11, NOVEMBER 2014

5940 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 13, NO. 11, NOVEMBER 2014 5940 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 13, NO. 11, NOVEMBER 014 Topoogy-Transparent Scheduing in Mobie Ad Hoc Networks With Mutipe Packet Reception Capabiity Yiming Liu, Member, IEEE,

More information

Endoscopic Motion Compensation of High Speed Videoendoscopy

Endoscopic Motion Compensation of High Speed Videoendoscopy Endoscopic Motion Compensation of High Speed Videoendoscopy Bharath avuri Department of Computer Science and Engineering, University of South Caroina, Coumbia, SC - 901. ravuri@cse.sc.edu Abstract. High

More information

On-Chip CNN Accelerator for Image Super-Resolution

On-Chip CNN Accelerator for Image Super-Resolution On-Chip CNN Acceerator for Image Super-Resoution Jung-Woo Chang and Suk-Ju Kang Dept. of Eectronic Engineering, Sogang University, Seou, South Korea {zwzang91, sjkang}@sogang.ac.kr ABSTRACT To impement

More information

Published in: Proc. IEEE International Conference on Acoustics, Speech and Signal Processing, April, 2003

Published in: Proc. IEEE International Conference on Acoustics, Speech and Signal Processing, April, 2003 Aaborg Universitet Compressed Domain Packet Loss Conceament of Sinusoiday Coded Speech Rødbro, Christoffer Asgaard; Christensen, Mads Græsbø; Andersen, Søren Vang; Jensen, Søren Hodt Pubished in: Proc.

More information

Application of Intelligence Based Genetic Algorithm for Job Sequencing Problem on Parallel Mixed-Model Assembly Line

Application of Intelligence Based Genetic Algorithm for Job Sequencing Problem on Parallel Mixed-Model Assembly Line American J. of Engineering and Appied Sciences 3 (): 5-24, 200 ISSN 94-7020 200 Science Pubications Appication of Inteigence Based Genetic Agorithm for Job Sequencing Probem on Parae Mixed-Mode Assemby

More information

A Comparison of a Second-Order versus a Fourth- Order Laplacian Operator in the Multigrid Algorithm

A Comparison of a Second-Order versus a Fourth- Order Laplacian Operator in the Multigrid Algorithm A Comparison of a Second-Order versus a Fourth- Order Lapacian Operator in the Mutigrid Agorithm Kaushik Datta (kdatta@cs.berkeey.edu Math Project May 9, 003 Abstract In this paper, the mutigrid agorithm

More information

NOWDAYS, our ways of living are under gradual but

NOWDAYS, our ways of living are under gradual but Transmission Management of Deay-Sensitive Medica Packets in Beyond Wireess Body Area Networks: A Queueing Game Approach Changyan Yi, Student Member, IEEE, and Jun Cai, Senior Member, IEEE Abstract In this

More information

Service Scheduling for General Packet Radio Service Classes

Service Scheduling for General Packet Radio Service Classes Service Scheduing for Genera Packet Radio Service Casses Qixiang Pang, Amir Bigoo, Victor C. M. Leung, Chris Schoefied Department of Eectrica and Computer Engineering, University of British Coumbia, Vancouver,

More information

Special Edition Using Microsoft Excel Selecting and Naming Cells and Ranges

Special Edition Using Microsoft Excel Selecting and Naming Cells and Ranges Specia Edition Using Microsoft Exce 2000 - Lesson 3 - Seecting and Naming Ces and.. Page 1 of 8 [Figures are not incuded in this sampe chapter] Specia Edition Using Microsoft Exce 2000-3 - Seecting and

More information

Community-Aware Opportunistic Routing in Mobile Social Networks

Community-Aware Opportunistic Routing in Mobile Social Networks IEEE TRANSACTIONS ON COMPUTERS VOL:PP NO:99 YEAR 213 Community-Aware Opportunistic Routing in Mobie Socia Networks Mingjun Xiao, Member, IEEE Jie Wu, Feow, IEEE, and Liusheng Huang, Member, IEEE Abstract

More information

Distance Weighted Discrimination and Second Order Cone Programming

Distance Weighted Discrimination and Second Order Cone Programming Distance Weighted Discrimination and Second Order Cone Programming Hanwen Huang, Xiaosun Lu, Yufeng Liu, J. S. Marron, Perry Haaand Apri 3, 2012 1 Introduction This vignette demonstrates the utiity and

More information

DETERMINING INTUITIONISTIC FUZZY DEGREE OF OVERLAPPING OF COMPUTATION AND COMMUNICATION IN PARALLEL APPLICATIONS USING GENERALIZED NETS

DETERMINING INTUITIONISTIC FUZZY DEGREE OF OVERLAPPING OF COMPUTATION AND COMMUNICATION IN PARALLEL APPLICATIONS USING GENERALIZED NETS DETERMINING INTUITIONISTIC FUZZY DEGREE OF OVERLAPPING OF COMPUTATION AND COMMUNICATION IN PARALLEL APPLICATIONS USING GENERALIZED NETS Pave Tchesmedjiev, Peter Vassiev Centre for Biomedica Engineering,

More information

Arithmetic Coding. Prof. Ja-Ling Wu. Department of Computer Science and Information Engineering National Taiwan University

Arithmetic Coding. Prof. Ja-Ling Wu. Department of Computer Science and Information Engineering National Taiwan University Arithmetic Coding Prof. Ja-Ling Wu Department of Computer Science and Information Engineering Nationa Taiwan University F(X) Shannon-Fano-Eias Coding W..o.g. we can take X={,,,m}. Assume p()>0 for a. The

More information

Research of Classification based on Deep Neural Network

Research of  Classification based on Deep Neural Network 2018 Internationa Conference on Sensor Network and Computer Engineering (ICSNCE 2018) Research of Emai Cassification based on Deep Neura Network Wang Yawen Schoo of Computer Science and Engineering Xi

More information

A Near-Optimal Distributed QoS Constrained Routing Algorithm for Multichannel Wireless Sensor Networks

A Near-Optimal Distributed QoS Constrained Routing Algorithm for Multichannel Wireless Sensor Networks Sensors 2013, 13, 16424-16450; doi:10.3390/s131216424 Artice OPEN ACCESS sensors ISSN 1424-8220 www.mdpi.com/journa/sensors A Near-Optima Distributed QoS Constrained Routing Agorithm for Mutichanne Wireess

More information

On Finding the Best Partial Multicast Protection Tree under Dual-Homing Architecture

On Finding the Best Partial Multicast Protection Tree under Dual-Homing Architecture On inding the est Partia Muticast Protection Tree under ua-homing rchitecture Mei Yang, Jianping Wang, Xiangtong Qi, Yingtao Jiang epartment of ectrica and omputer ngineering, University of Nevada Las

More information

BGP ingress-to-egress route configuration in a capacityconstrained Asia-Pacific Conference On Communications, 2005, v. 2005, p.

BGP ingress-to-egress route configuration in a capacityconstrained Asia-Pacific Conference On Communications, 2005, v. 2005, p. Tite BGP -to- route configuration in a capacityconstrained AS Author(s) Chim, TW; Yeung, KL; Lu KS Citation 2005 Asia-Pacific Conference On Communications, 2005, v. 2005, p. 386-390 Issued Date 2005 URL

More information

Self-Control Cyclic Access with Time Division - A MAC Proposal for The HFC System

Self-Control Cyclic Access with Time Division - A MAC Proposal for The HFC System Sef-Contro Cycic Access with Time Division - A MAC Proposa for The HFC System S.M. Jiang, Danny H.K. Tsang, Samue T. Chanson Hong Kong University of Science & Technoogy Cear Water Bay, Kowoon, Hong Kong

More information

Split Restoration with Wavelength Conversion in WDM Networks*

Split Restoration with Wavelength Conversion in WDM Networks* Spit Reoration with aveength Conversion in DM Networks* Yuanqiu Luo and Nirwan Ansari Advanced Networking Laborator Department of Eectrica and Computer Engineering New Jerse Initute of Technoog Universit

More information

Efficient method to design RF pulses for parallel excitation MRI using gridding and conjugate gradient

Efficient method to design RF pulses for parallel excitation MRI using gridding and conjugate gradient Origina rtice Efficient method to design RF puses for parae excitation MRI using gridding and conjugate gradient Shuo Feng, Jim Ji Department of Eectrica & Computer Engineering, Texas & M University, Texas,

More information

Ad Hoc Networks 11 (2013) Contents lists available at SciVerse ScienceDirect. Ad Hoc Networks

Ad Hoc Networks 11 (2013) Contents lists available at SciVerse ScienceDirect. Ad Hoc Networks Ad Hoc Networks (3) 683 698 Contents ists avaiabe at SciVerse ScienceDirect Ad Hoc Networks journa homepage: www.esevier.com/ocate/adhoc Dynamic agent-based hierarchica muticast for wireess mesh networks

More information

Quality Assessment using Tone Mapping Algorithm

Quality Assessment using Tone Mapping Algorithm Quaity Assessment using Tone Mapping Agorithm Nandiki.pushpa atha, Kuriti.Rajendra Prasad Research Schoar, Assistant Professor, Vignan s institute of engineering for women, Visakhapatnam, Andhra Pradesh,

More information

A Fast-Convergence Decoding Method and Memory-Efficient VLSI Decoder Architecture for Irregular LDPC Codes in the IEEE 802.

A Fast-Convergence Decoding Method and Memory-Efficient VLSI Decoder Architecture for Irregular LDPC Codes in the IEEE 802. A Fast-Convergence Decoding Method and Memory-Efficient VLSI Decoder Architecture for Irreguar LDPC Codes in the IEEE 82.16e Standards Yeong-Luh Ueng and Chung-Chao Cheng Dept. of Eectrica Engineering,

More information

Testing Whether a Set of Code Words Satisfies a Given Set of Constraints *

Testing Whether a Set of Code Words Satisfies a Given Set of Constraints * JOURNAL OF INFORMATION SCIENCE AND ENGINEERING 6, 333-346 (010) Testing Whether a Set of Code Words Satisfies a Given Set of Constraints * HSIN-WEN WEI, WAN-CHEN LU, PEI-CHI HUANG, WEI-KUAN SHIH AND MING-YANG

More information

file://j:\macmillancomputerpublishing\chapters\in073.html 3/22/01

file://j:\macmillancomputerpublishing\chapters\in073.html 3/22/01 Page 1 of 15 Chapter 9 Chapter 9: Deveoping the Logica Data Mode The information requirements and business rues provide the information to produce the entities, attributes, and reationships in ogica mode.

More information

PAPER Delay Constrained Routing and Link Capacity Assignment in Virtual Circuit Networks

PAPER Delay Constrained Routing and Link Capacity Assignment in Virtual Circuit Networks 2004 PAPER Deay Constrained Routing and Link Capacity Assignment in Virtua Circuit Networks Hong-Hsu YEN a), Member and FrankYeong-Sung LIN b), Nonmember SUMMARY An essentia issue in designing, operating

More information

Delay Budget Partitioning to Maximize Network Resource Usage Efficiency

Delay Budget Partitioning to Maximize Network Resource Usage Efficiency Deay Budget Partitioning to Maximize Network Resource Usage Efficiency Kartik Gopaan Tzi-cker Chiueh Yow-Jian Lin Forida State University Stony Brook University Tecordia Technoogies kartik@cs.fsu.edu chiueh@cs.sunysb.edu

More information

Path-Based Protection for Surviving Double-Link Failures in Mesh-Restorable Optical Networks

Path-Based Protection for Surviving Double-Link Failures in Mesh-Restorable Optical Networks Path-Based Protection for Surviving Doube-Link Faiures in Mesh-Restorabe Optica Networks Wensheng He and Arun K. Somani Dependabe Computing and Networking Laboratory Department of Eectrica and Computer

More information

Multiple Plane Phase Retrieval Based On Inverse Regularized Imaging and Discrete Diffraction Transform

Multiple Plane Phase Retrieval Based On Inverse Regularized Imaging and Discrete Diffraction Transform Mutipe Pane Phase Retrieva Based On Inverse Reguaried Imaging and Discrete Diffraction Transform Artem Migukin, Vadimir Katkovnik, and Jaakko Astoa Department of Signa Processing, Tampere University of

More information

As Michi Henning and Steve Vinoski showed 1, calling a remote

As Michi Henning and Steve Vinoski showed 1, calling a remote Reducing CORBA Ca Latency by Caching and Prefetching Bernd Brügge and Christoph Vismeier Technische Universität München Method ca atency is a major probem in approaches based on object-oriented middeware

More information

FIRST BEZIER POINT (SS) R LE LE. φ LE FIRST BEZIER POINT (PS)

FIRST BEZIER POINT (SS) R LE LE. φ LE FIRST BEZIER POINT (PS) Singe- and Muti-Objective Airfoi Design Using Genetic Agorithms and Articia Inteigence A.P. Giotis K.C. Giannakogou y Nationa Technica University of Athens, Greece Abstract Transonic airfoi design probems

More information

Image Segmentation Using Semi-Supervised k-means

Image Segmentation Using Semi-Supervised k-means I J C T A, 9(34) 2016, pp. 595-601 Internationa Science Press Image Segmentation Using Semi-Supervised k-means Reza Monsefi * and Saeed Zahedi * ABSTRACT Extracting the region of interest is a very chaenging

More information

An Exponential Time 2-Approximation Algorithm for Bandwidth

An Exponential Time 2-Approximation Algorithm for Bandwidth An Exponentia Time 2-Approximation Agorithm for Bandwidth Martin Fürer 1, Serge Gaspers 2, Shiva Prasad Kasiviswanathan 3 1 Computer Science and Engineering, Pennsyvania State University, furer@cse.psu.edu

More information

QoS-Aware Data Transmission and Wireless Energy Transfer: Performance Modeling and Optimization

QoS-Aware Data Transmission and Wireless Energy Transfer: Performance Modeling and Optimization QoS-Aware Data Transmission and Wireess Energy Transfer: Performance Modeing and Optimization Dusit Niyato, Ping Wang, Yeow Wai Leong, and Tan Hwee Pink Schoo of Computer Engineering, Nanyang Technoogica

More information

Resource Optimization to Provision a Virtual Private Network Using the Hose Model

Resource Optimization to Provision a Virtual Private Network Using the Hose Model Resource Optimization to Provision a Virtua Private Network Using the Hose Mode Monia Ghobadi, Sudhakar Ganti, Ghoamai C. Shoja University of Victoria, Victoria C, Canada V8W 3P6 e-mai: {monia, sganti,

More information

Formulation of Loss minimization Problem Using Genetic Algorithm and Line-Flow-based Equations

Formulation of Loss minimization Problem Using Genetic Algorithm and Line-Flow-based Equations Formuation of Loss minimization Probem Using Genetic Agorithm and Line-Fow-based Equations Sharanya Jaganathan, Student Member, IEEE, Arun Sekar, Senior Member, IEEE, and Wenzhong Gao, Senior member, IEEE

More information

Real-Time Image Generation with Simultaneous Video Memory Read/Write Access and Fast Physical Addressing

Real-Time Image Generation with Simultaneous Video Memory Read/Write Access and Fast Physical Addressing Rea-Time Image Generation with Simutaneous Video Memory Read/rite Access and Fast Physica Addressing Mountassar Maamoun 1, Bouaem Laichi 2, Abdehaim Benbekacem 3, Daoud Berkani 4 1 Department of Eectronic,

More information

MCSE Training Guide: Windows Architecture and Memory

MCSE Training Guide: Windows Architecture and Memory MCSE Training Guide: Windows 95 -- Ch 2 -- Architecture and Memory Page 1 of 13 MCSE Training Guide: Windows 95-2 - Architecture and Memory This chapter wi hep you prepare for the exam by covering the

More information

Space-Time Trade-offs.

Space-Time Trade-offs. Space-Time Trade-offs. Chethan Kamath 03.07.2017 1 Motivation An important question in the study of computation is how to best use the registers in a CPU. In most cases, the amount of registers avaiabe

More information

A Method for Calculating Term Similarity on Large Document Collections

A Method for Calculating Term Similarity on Large Document Collections $ A Method for Cacuating Term Simiarity on Large Document Coections Wofgang W Bein Schoo of Computer Science University of Nevada Las Vegas, NV 915-019 bein@csunvedu Jeffrey S Coombs and Kazem Taghva Information

More information

power-saving mode for mobile computing in Wi-Fi hotspots: Limitations, enhancements and open issues

power-saving mode for mobile computing in Wi-Fi hotspots: Limitations, enhancements and open issues DOI 10.1007/s11276-006-0010-9 802.11 power-saving mode for mobie computing in Wi-Fi hotspots: Limitations, enhancements and open issues G. Anastasi M. Conti E. Gregori A. Passarea C Science + Business

More information

A HIGH PERFORMANCE, LOW LATENCY, LOW POWER AUDIO PROCESSING SYSTEM FOR WIDEBAND SPEECH OVER WIRELESS LINKS

A HIGH PERFORMANCE, LOW LATENCY, LOW POWER AUDIO PROCESSING SYSTEM FOR WIDEBAND SPEECH OVER WIRELESS LINKS A HIGH PERFORMANCE, LOW LATENCY, LOW POWER AUDIO PROCESSING SYSTEM FOR WIDEBAND SPEECH OVER WIRELESS LINKS Etienne Cornu 1, Aain Dufaux 2, and David Hermann 1 1 AMI Semiconductor Canada, 611 Kumpf Drive,

More information

Intro to Programming & C Why Program? 1.2 Computer Systems: Hardware and Software. Why Learn to Program?

Intro to Programming & C Why Program? 1.2 Computer Systems: Hardware and Software. Why Learn to Program? Intro to Programming & C++ Unit 1 Sections 1.1-3 and 2.1-10, 2.12-13, 2.15-17 CS 1428 Spring 2018 Ji Seaman 1.1 Why Program? Computer programmabe machine designed to foow instructions Program a set of

More information

Joint Optimization of Intra- and Inter-Autonomous System Traffic Engineering

Joint Optimization of Intra- and Inter-Autonomous System Traffic Engineering Joint Optimization of Intra- and Inter-Autonomous System Traffic Engineering Kin-Hon Ho, Michae Howarth, Ning Wang, George Pavou and Styianos Georgouas Centre for Communication Systems Research, University

More information

1682 IEEE TRANSACTIONS ON FUZZY SYSTEMS, VOL. 22, NO. 6, DECEMBER Backward Fuzzy Rule Interpolation

1682 IEEE TRANSACTIONS ON FUZZY SYSTEMS, VOL. 22, NO. 6, DECEMBER Backward Fuzzy Rule Interpolation 1682 IEEE TRANSACTIONS ON FUZZY SYSTEMS, VOL. 22, NO. 6, DECEMBER 2014 Bacward Fuzzy Rue Interpoation Shangzhu Jin, Ren Diao, Chai Que, Senior Member, IEEE, and Qiang Shen Abstract Fuzzy rue interpoation

More information

Fastest-Path Computation

Fastest-Path Computation Fastest-Path Computation DONGHUI ZHANG Coege of Computer & Information Science Northeastern University Synonyms fastest route; driving direction Definition In the United states, ony 9.% of the househods

More information

A New Supervised Clustering Algorithm Based on Min-Max Modular Network with Gaussian-Zero-Crossing Functions

A New Supervised Clustering Algorithm Based on Min-Max Modular Network with Gaussian-Zero-Crossing Functions 2006 Internationa Joint Conference on Neura Networks Sheraton Vancouver Wa Centre Hote, Vancouver, BC, Canada Juy 16-21, 2006 A New Supervised Custering Agorithm Based on Min-Max Moduar Network with Gaussian-Zero-Crossing

More information

On coding for reliable communication over packet networks

On coding for reliable communication over packet networks Physica Communication 1 (2008) 3 20 Fu ength artice www.esevier.com/ocate/phycom On coding for reiabe communication over packet networks Desmond S. Lun a,1, Murie Médard a,, Raf Koetter b,2, Michee Effros

More information

Utility-based Camera Assignment in a Video Network: A Game Theoretic Framework

Utility-based Camera Assignment in a Video Network: A Game Theoretic Framework This artice has been accepted for pubication in a future issue of this journa, but has not been fuy edited. Content may change prior to fina pubication. Y.LI AND B.BHANU CAMERA ASSIGNMENT: A GAME-THEORETIC

More information

Reference trajectory tracking for a multi-dof robot arm

Reference trajectory tracking for a multi-dof robot arm Archives of Contro Sciences Voume 5LXI, 5 No. 4, pages 53 57 Reference trajectory tracking for a muti-dof robot arm RÓBERT KRASŇANSKÝ, PETER VALACH, DÁVID SOÓS, JAVAD ZARBAKHSH This paper presents the

More information

CERIAS Tech Report Replicated Parallel I/O without Additional Scheduling Costs by Mikhail J. Atallah Center for Education and Research

CERIAS Tech Report Replicated Parallel I/O without Additional Scheduling Costs by Mikhail J. Atallah Center for Education and Research CERIAS Tech Report 2003-50 Repicated Parae I/O without Additiona Scheduing Costs by Mikhai J. Ataah Center for Education and Research Information Assurance and Security Purdue University, West Lafayette,

More information

Automatic Grouping for Social Networks CS229 Project Report

Automatic Grouping for Social Networks CS229 Project Report Automatic Grouping for Socia Networks CS229 Project Report Xiaoying Tian Ya Le Yangru Fang Abstract Socia networking sites aow users to manuay categorize their friends, but it is aborious to construct

More information

Crossing Minimization Problems of Drawing Bipartite Graphs in Two Clusters

Crossing Minimization Problems of Drawing Bipartite Graphs in Two Clusters Crossing Minimiation Probems o Drawing Bipartite Graphs in Two Custers Lanbo Zheng, Le Song, and Peter Eades Nationa ICT Austraia, and Schoo o Inormation Technoogies, University o Sydney,Austraia Emai:

More information

Cross-layer Design for Efficient Resource Utilization in WiMedia UWB-based WPANs

Cross-layer Design for Efficient Resource Utilization in WiMedia UWB-based WPANs Cross-ayer Design for Efficient Resource Utiization in WiMedia UWB-based WPANs RAED AL-ZUBI and MARWAN KRUNZ Department of Eectrica and Computer Engineering. University of Arizona. Utra-wideband (UWB)

More information

Providing Hop-by-Hop Authentication and Source Privacy in Wireless Sensor Networks

Providing Hop-by-Hop Authentication and Source Privacy in Wireless Sensor Networks The 31st Annua IEEE Internationa Conference on Computer Communications: Mini-Conference Providing Hop-by-Hop Authentication and Source Privacy in Wireess Sensor Networks Yun Li Jian Li Jian Ren Department

More information

TSR: Topology Reduction from Tree to Star Data Grids

TSR: Topology Reduction from Tree to Star Data Grids 03 Seventh Internationa Conference on Innovative Mobie and Internet Services in biquitous Computing TSR: Topoogy Reduction from Tree to Star Data Grids Ming-Chang Lee #, Fang-Yie Leu *, Ying-ping Chen

More information

Understanding the Mixing Patterns of Social Networks: The Impact of Cores, Link Directions, and Dynamics

Understanding the Mixing Patterns of Social Networks: The Impact of Cores, Link Directions, and Dynamics Understanding the Mixing Patterns of Socia Networks: The Impact of Cores, Link Directions, and Dynamics [Last revised on May 22, 2011] Abedeaziz Mohaisen Huy Tran Nichoas Hopper Yongdae Kim University

More information

Adaptive 360 VR Video Streaming: Divide and Conquer!

Adaptive 360 VR Video Streaming: Divide and Conquer! Adaptive 360 VR Video Streaming: Divide and Conquer! Mohammad Hosseini *, Viswanathan Swaminathan * University of Iinois at Urbana-Champaign (UIUC) Adobe Research, San Jose, USA Emai: shossen2@iinois.edu,

More information

CSE120 Principles of Operating Systems. Prof Yuanyuan (YY) Zhou Scheduling

CSE120 Principles of Operating Systems. Prof Yuanyuan (YY) Zhou Scheduling CSE120 Principes of Operating Systems Prof Yuanyuan (YY) Zhou Scheduing Announcement Homework 2 due on October 25th Project 1 due on October 26th 2 CSE 120 Scheduing and Deadock Scheduing Overview In discussing

More information

Layout Conscious Approach and Bus Architecture Synthesis for Hardware-Software Co-Design of Systems on Chip Optimized for Speed

Layout Conscious Approach and Bus Architecture Synthesis for Hardware-Software Co-Design of Systems on Chip Optimized for Speed Layout Conscious Approach and Bus Architecture Synthesis for Hardware-Software Co-Design of Systems on Chip Optimized for Speed Nattawut Thepayasuwan, Member, IEEE and Aex Doboi, Member, IEEE Abstract

More information

Research on the overall optimization method of well pattern in water drive reservoirs

Research on the overall optimization method of well pattern in water drive reservoirs J Petro Expor Prod Techno (27) 7:465 47 DOI.7/s322-6-265-3 ORIGINAL PAPER - EXPLORATION ENGINEERING Research on the overa optimization method of we pattern in water drive reservoirs Zhibin Zhou Jiexiang

More information

CORRELATION filters (CFs) are a useful tool for a variety

CORRELATION filters (CFs) are a useful tool for a variety Zero-Aiasing Correation Fiters for Object Recognition Joseph A. Fernandez, Student Member, IEEE, Vishnu Naresh Boddeti, Member, IEEE, Andres Rodriguez, Member, IEEE, B. V. K. Vijaya Kumar, Feow, IEEE arxiv:4.36v

More information

Peer-Assisted Computation Offloading in Wireless Networks, Student Member, IEEE, and Guohong Cao, Fellow, IEEE

Peer-Assisted Computation Offloading in Wireless Networks, Student Member, IEEE, and Guohong Cao, Fellow, IEEE IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 17, NO. 7, JULY 2018 4565 Peer-Assisted Computation Offoading in Wireess Networks Yei Geng, Student Member, IEEE, and Guohong Cao, Feow, IEEE Abstract

More information

Alpha labelings of straight simple polyominal caterpillars

Alpha labelings of straight simple polyominal caterpillars Apha abeings of straight simpe poyomina caterpiars Daibor Froncek, O Nei Kingston, Kye Vezina Department of Mathematics and Statistics University of Minnesota Duuth University Drive Duuth, MN 82-3, U.S.A.

More information