ECE Department, Carnegie Mellon University, Pittsburgh, PA

Size: px
Start display at page:

Download "ECE Department, Carnegie Mellon University, Pittsburgh, PA"

Transcription

1 Bayesian odel Fusion: Large-Scale Perforance odeling of Analog and ixed-signal Circuits by Reusing Early-Stage Data Fa Wang, Wangyang Zhang, Shupeng Sun, Xin Li and Chenjie Gu ECE Departent, Carnegie ellon University, Pittsburgh, PA 5 Strategic CAD Labs, Intel Corporation, Hillsboro, OR 974 {fwang, wyzhang, shupengs, xinli}@ece.cu.edu, chenjie.gu@intel.co ABSTRACT Efficient high-diensional perforance odeling of today s coplex analog and ixed-signal (AS circuits with large-scale process variations is an iportant yet challenging task. In this paper, we propose a novel perforance odeling algorith that is referred to as Bayesian odel Fusion (BF. Our key idea is to borrow the siulation data generated fro an early stage (e.g., scheatic level to facilitate efficient high-diensional perforance odeling at a late stage (e.g., post layout with low coputational cost. Such a goal is achieved by statistically odeling the perforance correlation between early and late stages through Bayesian inference. Several circuit exaples designed in a coercial n COS process deonstrate that BF achieves up to 9 runtie speedup over the traditional odeling technique without surrendering any accuracy.. INTRODUCTION The aggressive scaling of integrated circuits (ICs leads to large-scale process variations that cannot be easily reduced by foundries. Process variations anifest theselves as the uncertainties associated with the geoetrical and electrical paraeters of seiconductor devices. These device-level variations significantly ipact the paraetric yield of analog and ixed-signal (AS circuits and, hence, ust be appropriately odeled, analyzed and optiized at all levels of design hierarchy []-[]. To address this variability issue, various techniques for perforance odeling have been developed during the past two decades []-[8]. The objective is to approxiate the circuit-level perforance (e.g., gain of an analog aplifier as an analytical (e.g., linear, quadratic, etc function of device-level variations (e.g., ΔV TH, ΔT OX, etc. Once such a perforance odel is available, it can be applied to a nuber of iportant applications such as estiating paraetric yield [9], extracting worst-case corner [], optiizing circuit design []-[5], etc. While perforance odeling was extensively studied in the past, the evolution of today s AS circuits has posed a nuber of new challenges in this area. In particular, the recent adoption of several eerging design ethodologies (e.g., reconfigurable analog design, adaptive post-silicon tuning, etc leads to highly coplex AS systes that integrate nuerous nanoscale devices. The rearkable increase of AS circuit size results in a two-fold consequence. Perission to ake digital or hard copies of all or part of this work for personal or classroo use is granted without fee provided that copies are not ade or distributed for profit or coercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific perission and/or a fee. DAC, ay 9 - June 7, Austin, TX, USA. Copyright AC //5...$5. High-diensional variation space: A large nuber of devicelevel rando variables ust be used to odel the process variations associated with a large-scale AS syste. For exaple, about 4 independent rando variables are required to odel the device isatches of a single transistor for a coercial n COS process. If an AS syste contains 4 transistors, there are about 4 5 rando variables in total to capture the corresponding device-level variations, resulting in a high-diensional variation space. In addition, it is extreely difficult, if not ipossible, to pre-select a subset of these rando variables for variation analysis, since the ipact of device isatches is circuit- and perforance-dependent. Expensive circuit siulation: The coputational cost of circuit siulation substantially increases, as the AS circuit size becoes increasingly large. For instance, it ay take a few days or even a few weeks to run the transistor-level siulation of a large AS circuit such as phase-locked loop or high-speed link. These recent trends of today s AS circuits ake perforance odeling extreely difficult. On one hand, a large nuber of siulation saples ust be generated in order to fit a high-diensional odel. On the other hand, creating a single sapling point by transistor-level siulation can take a large aount of coputational tie. The challenging issue here is how to ake perforance odeling coputationally affordable for today s large-scale AS circuits. This fundaental issue has not been appropriately addressed by the traditional perforance odeling techniques, e.g., the recent sparse regression algorith based on Orthogonal atching Pursuit (OP [8]. In this paper, we propose a new Bayesian odel Fusion (BF technique to facilitate large-scale perforance odeling of AS circuits. The proposed BF ethod is otivated by the fact that today s AS circuits are often designed via a ulti-stage flow. Naely, an AS design often spans three core stages: (i scheatic design, (ii layout design, and (iii chip anufacturing and testing. At each stage, siulation or easureent data are collected to validate the circuit design, before oving to the next stage. The traditional perforance odeling techniques rely on the data at a single stage only and they copletely ignore the data that are generated at other stages. The key idea of BF, however, is to reuse the early-stage data when fitting a late-stage perforance odel. As such, the perforance odeling cost can be substantially reduced. atheatically, the proposed BF ethod is derived fro the theory of Bayesian inference. Starting fro a set of early-stage (e.g., scheatic-level sapling points, BF first approxiates an early-stage perforance odel based on these saples. The early-stage odel is used as a teplate to define our prior knowledge for late-stage (e.g., post-layout perforance odeling. Specifically, a prior distribution is statistically defined for the late-stage odel coefficients. The prior knowledge is then cobined with very few late-stage sapling points to solve the

2 late-stage odel coefficients via Bayesian inference. Fro this point of view, by fusing the early-stage and late-stage perforance odels through Bayesian inference, we only need a sall nuber of late-stage sapling points to fit a highdiensional late-stage odel, thereby significantly reducing the coputational cost for perforance odeling. As will be deonstrated by our nuerical exaples in Section 5, BF achieves up to 9 runtie speedup over the traditional odeling technique without surrendering any accuracy. BF was previously proposed for paraetric yield estiation of AS circuits in [6] where Bayesian inference was used to estiate the probability distribution of AS perforance etrics. In this paper, we further extend the idea of BF to perforance odeling. It is iportant to ephasize that the forulation of our prior knowledge for perforance odeling is copletely different fro that shown in [6], as will be discussed in Section. The reainder of this paper is organized as follows. In Section, we review the iportant background on perforance odeling, and then derive our proposed BF ethod in Section. Several ipleentation issues are discussed in Section 4. The efficacy of BF is deonstrated by several circuit exaples in Section 5. Finally, we conclude in Section 6.. BACGROUND Given an AS circuit (e.g., an analog aplifier, its perforance (e.g., gain ay vary due to device-level variations (e.g., ΔV TH, ΔT OX, etc. The objective of perforance odeling is to approxiate the circuit perforance as an analytical function of the device-level variations: ( ( x f = x α ( g where f represents the perforance of interest, x is a vector containing the rando variables to odel device-level variations, {α ; =,,..., } denote the odel coefficients, {g (x; =,,..., } are the basis functions (e.g., linear or quadratic polynoials, and is the total nuber of basis functions. In order to deterine the perforance odel in (, we need to find the odel coefficients {α ; =,,..., }. Towards this goal, the traditional least-squares fitting ethod first generates a set of sapling points and then solves the odel coefficients fro the following linear equation [8]: G α = f ( where ( ( ( ( ( ( g x g x g x ( ( ( ( ( ( g x g x g x G = ( ( ( ( ( ( ( g x g x g x α α (4 5 ( ( [ ( f = f f f ] T. (5 In (-(5, x (k and f (k are the values of x and f(x at the kth sapling point respectively, and represents the total nuber of sapling points. The nuber of sapling points (i.e., should be greater than the nuber of unknown coefficients (i.e.,. As such, the linear equation in ( is overdeterined and the unknown odel coefficients {α ; =,,..., } are found by solving its least-squares solution. When the aforeentioned least-squares fitting ethod is applied to fit a high-diensional perforance odel with any 4 = [ α α ] T unknown odel coefficients, it requires a large nuber of sapling points to for the overdeterined linear equation in (. Note that each sapling point is generated by running an expensive transistor-level siulation. It, in turn, iplies that the least-squares fitting approach can be extreely expensive for high-diensional perforance odeling. Recently, sparse regression has been developed to address this coplexity issue [8]. The key idea is not to solve an overdeterined linear equation. Instead, the unknown odel coefficients are uniquely deterined by solving an underdeterined linear equation. This goal is achieved by exploiting the fact that ost odel coefficients of a highdiensional perforance odel are close to zero. In other words, the unknown odel coefficients carry a unique sparse pattern. The sparse regression algoriths were particularly developed to solve these sparse coefficients fro a sall nuber of sapling points. As such, the siulation cost of generating the required sapling points is greatly reduced. While sparse regression has been successfully applied to any practical applications, it still requires a large nuber of (e.g., sapling points to fit a high-diensional perforance odel [8]. Therefore, it reains ill-equipped for odeling largescale AS circuits where running a single transistor-level siulation to generate one sapling point ay take a few days or even a few weeks. otivated by this observation, we will propose a new BF technique in this paper to further reduce the nuber of required siulation saples and, hence, the coputational cost for large-scale perforance odeling.. BAYESIAN ODEL FUSION Siilar to sparse regression, the proposed BF ethod relies on the assuption that ost odel coefficients of a highdiensional perforance odel are close to zero. However, unlike the traditional sparse regression approach that fits the sparse perforance odel based on the siulation data at a single stage only (e.g., post-layout siulation data, BF attepts to identify the underlying sparse pattern by re-using the early-stage data (e.g., scheatic-level siulation data in order to efficiently fit a late-stage (e.g., post-layout perforance odel. In particular, BF consists of the following two ajor steps: (i statistically defining the prior knowledge of the sparse pattern based on the early-stage siulation data, and (ii optially deterining the late-stage perforance odel by cobining the prior knowledge and very few late-stage siulation saples. In this section, we will discuss the atheatical forulation of these two steps and highlight the novelty.. Prior nowledge Definition We consider two different perforance odels: the earlystage odel f E (x and the late-stage odel f L (x: 6 f E ( g ( x α (6 = E, x 7 f L ( g ( x α (7 = L, x where {α E, ; =,,..., } and {α L, ; =,,..., } represent the early-stage and late-stage odel coefficients, respectively. In (6-(7, we assue that the early-stage odel f E (x and the latestage odel f L (x share the sae basis functions. ore coplicated cases where f E (x and f L (x are approxiated by different basis functions will be further discussed in Section 4..

3 The early-stage odel f E (x is fitted fro the early-stage siulation data. In practice, the early-stage siulation data are collected to validate the early-stage design, before we ove to the next stage. For this reason, we should already know the earlystage odel f E (x before fitting the late-stage odel f L (x. Naely, we assue that the early-stage odel coefficients {α E, ; =,,..., } are provided as the input to our proposed BF ethod for late-stage perforance odeling. Given the early-stage odel f E (x, we first extract the prior knowledge that can be used to facilitate efficient late-stage odeling. To this end, we propose to learn the underlying sparse pattern for the late-stage odel f L (x based on the early-stage odel coefficients {α E, ; =,,..., }. Reeber that both the early-stage and late-stage odels are fitted for the sae perforance etric of the sae circuit. Their odel coefficients should be siilar. Naely, if the early-stage odel coefficient α E, has a large (or sall agnitude, it is likely that the late-stage odel coefficient α L, also has a large (or sall agnitude. Such prior knowledge should be atheatically encoded into our proposed perforance odeling flow. In this paper, we statistically represent the prior knowledge as a probability density function (PDF that is referred to as the prior distribution [9]. In particular, we odel each late-stage odel coefficient as a zero-ean Gaussian distribution: 8 pdf ( exp α L, α = ~ N(, σ L, π σ ( =,,, σ where the standard deviation σ is a paraeter that encodes the agnitude inforation of the odel coefficient α L,. If the standard deviation σ is sall, the prior distribution pdf(α L, is narrowly peaked around zero, iplying that the coefficient α L, is possibly close to zero. Otherwise, if the standard deviation σ is large, the prior distribution pdf(α L, widely spreads over a large range and the coefficient α L, can possibly take a value that is far away fro zero. Figure shows a siple exaple of our proposed prior distribution for two odel coefficients α L, and α L, where σ is sall and σ is large. pdf(α L, ~ N(, σ PDF pdf(α L, ~ N(, σ α L, or α L, Figure. A siple exaple of our proposed prior distribution is shown for two odel coefficients α L, and α L,. The coefficient α L, is possibly close to zero, since its prior distribution is narrowly peaked around zero. The coefficient α L, can possibly be far away fro zero, since its prior distribution widely spreads over a large range. Given (8, we need to appropriately deterine the standard deviation σ to fully specify the prior distribution pdf(α L,. The value of σ should be optiized so that the probability distribution pdf(α L, correctly represents our prior knowledge. In other words, by appropriately choosing the value of σ, the prior distribution pdf(α L, should take a large value (i.e., a high probability at the location where the actual late-stage odel coefficient α L, occurs. However, we only know the early-stage odel coefficient α E,, instead of the late-stage odel coefficient (8 α L,, at this oent. Reeber that α E, and α L, are expected to be siilar. Hence, the prior distribution pdf(α L, should also take a large value at α L, = α E,. Based on this criterion, the optial prior distribution pdf(α L, can be found by axiizing the probability for α E, to occur: 9 ax pdf ( α L, = α E, ( =,,,. (9 σ Naely, given the early-stage odel coefficient α E,, the optial standard deviation σ is deterined by the axiu likelihood estiation (LE in (9. To solve σ fro (9, we consider the following first-order optiality condition: d pdf ( α L, = α E, = ( =,,,. ( dσ Substituting (8 into ( yields: E E exp α, α, = π σ σ σ σ. ( ( =,,, The optial value of σ is equal to: = α ( =,, σ. ( E,, Eq. ( reveals an iportant fact that the optial standard deviation σ is siply equal to the absolute value of the earlystage odel coefficient α E,. This observation is consistent with our intuition. Naely, if the early-stage odel coefficient α E, has a large (or sall agnitude, the late-stage odel coefficient α L, should also have a large (or sall agnitude and, hence, the standard deviation σ should be large (or sall. To coplete the definition of the prior distribution for all latestage odel coefficients {α L, ; =,,..., }, we further assue that these coefficients are statistically independent and their joint distribution is represented as: ( α L, pdf α = L exp ( = σ π σ where ( = 4 [ ] T L = α L, α L, α L, α (4 contains all late-stage odel coefficients. The independence assuption in ( siply iplies that we do not know the correlation inforation aong these coefficients as our prior knowledge. The correlation inforation will be learned fro the late-stage siulation data, when the posterior distribution is calculated by the Bayesian inference in Section.. Finally, it is iportant to ention that the prior knowledge can be possibly defined as a distribution that is different fro (8. For exaple, the prior distribution is specified as a Gaussian distribution with non-zero ean in [6]. It, however, does not encode the sparse pattern of odel coefficients, as is the case of this paper. The efficacy of different prior definitions is casedependent. It reains an open question how to deterine the optial prior distribution for a specific perforance odeling proble where the circuit and perforance of interest are given. This proble will be further studied in our future research.. axiu-a-posteriori Estiation Once the prior distribution pdf(α L is derived in (, we will cobine pdf(α L with late-stage siulation saples {(x (k, f L (k ; k =,,, }, where x (k and f L (k are the values of x and f L (x at

4 the kth sapling point respectively, to solve the late-stage odel coefficient α L by axiu-a-posteriori (AP estiation. The key idea of AP is to find the posterior distribution [9], i.e., the conditional PDF pdf(α L f L where 5 ( ( [ ( f ] T L = f L f L f L (5 contains all late-stage siulation saples that are collected. Intuitively, the posterior distribution pdf(α L f L indicates the reaining uncertainty of α L, after we observe late-stage siulation saples. Here, since α L is a rando variable, it is described by a probability distribution, instead of a deterinistic value. AP attepts to find the optial value of α L to axiize the posterior distribution pdf(α L f L. Naely, it ais to find the solution α L that is ost likely to occur according to the posterior distribution. Based on Bayes theore, the posterior distribution pdf(α L f L is proportional to the prior distribution pdf(α L ultiplied by the likelihood function pdf(f L α L [9]: 6 pdf ( α L f L pdf ( α L pdf ( f L α L. (6 The prior distribution pdf(α L is already defined in (. To derive the likelihood function pdf(f L α L, we further assue that the error for the late-stage perforance odel f L (x follows a zeroean Gaussian distribution and, hence, the approxiate equality in (7 can be re-written as: 7 f L ( x = α L, g ( x + ε L (7 = where ε L denotes the odeling error with the distribution: ε L 8 pdf ( ε L exp ~ N (, σ π σ σ =. (8 In (8, the standard deviation σ controls the agnitude of the odeling error. Its value can be optially deterined by using the cross-validation technique that will be discussed in Section 4.. Given (7-(8, since the odeling error at the kth siulation saple (x (k, f (k L is siply one sapling point of the rando variable ε L, it follows the Gaussian distribution: 9 ( ( ( k ( k f α g x ~ N σ. (9 L = L,, Therefore, the probability of observing the kth sapling point is: ( ( ( k k exp α, fl L g x σ ( ( = k pdf f = L α L.( π σ Assue that all sapling points are independently generated, we can write the likelihood function pdf(f L α L as: ( α L ( ( k pdf = L α L pdf f L f. ( k= Cobining (, (6 and (-(, it is straightforward to prove that the posterior distribution pdf(α L f L is Gaussian and its covariance atrix Σ L and ean vector μ L are [7], [9]: T Σ [ ( ] L = σ G G + diag σ, σ,, σ ( T μ L = σ Σ L G f L ( where G and f L are defined by ( and (5 respectively, and diag( represents an operator to construct a diagonal atrix. Since the Gaussian PDF pdf(α L f L reaches its axiu at the ean value, the AP solution α L is equal to the ean vector μ L : T 4 α L = σ Σ L G f L. (4 In other words, Eq. (4 shows the optial coefficients solved by our proposed BF ethod for the late-stage perforance odel f L (x. While the basic idea of prior knowledge definition and axiu-a-posteriori estiation is illustrated in this section, several ipleentation issues ust be carefully considered in order to ake BF of practical utility. These ipleentation details will be further discussed in the next section. 4. IPLEENTATION ISSUES To ake the proposed BF ethod of practical utility, two ipleentation issues, (i issing prior knowledge and (ii crossvalidation, ust be carefully considered. In this section, we will discuss these ipleentation issues in detail. 4. issing Prior nowledge The BF ethod derived in Section assues that the earlystage odel f E (x and the late-stage odel f L (x share the sae basis functions. In practice, this assuption ay not always hold, because the early-stage odel does not necessarily capture all the detailed behaviors of a circuit. For instance, it is well-known that layout parasitics will be added to the post-layout netlist (late stage during layout extraction. The variations of these parasitics ust be odeled by a nuber of new rando variables that are copletely ignored at the scheatic level (early stage. The latestage post-layout odel f L (x should contain additional basis functions corresponding to the new rando variables that are not found fro the early-stage scheatic odel f E (x. In this case, the early-stage odel f E (x does not carry any prior knowledge about the late-stage odel coefficients associated with these additional basis functions. In other words, the prior knowledge for these latestage odel coefficients is issing. To appropriately handle the cases with issing prior knowledge, we re-visit the prior distribution pdf(α L, defined in (8. As entioned in Section., the standard deviation σ of the Gaussian distribution pdf(α L, encodes the agnitude inforation of the late-stage odel coefficient α L,. If there is no prior knowledge available for α L,, it iplies that the late-stage odel coefficient α L, can possibly take any value with equal probability. Hence, the standard deviation σ should be set to + : 5 σ = + (5 so that the prior distribution is nearly constant over a wide range. Note that when calculating the posterior distribution in (-(, only the value of σ is needed. Hence, the infinite standard deviation in (5 would not cause any nuerical proble for solving the late-stage odel coefficients. 4. Cross-Validation As entioned in Section., the standard deviation σ of the odeling error in (8 ust be deterined. Otherwise, without knowing σ, the late-stage odel coefficients α L cannot be deterined by the AP solution in (4. The objective here is to find the optial value of σ so that the odeling error is iniized. Towards this goal, we ust accurately estiate the odeling error for different σ values and then select the optial σ with inial error. To quantitatively estiate the odeling error for a given σ value, we adopt the idea of N-fold cross validation fro the statistics counity [9]. Naely, we partition the entire data set 4

5 into N groups. odeling error is estiated fro N independent runs. In each run, one of the N groups is used to estiate the odeling error and all other groups are used to calculate the odel coefficients. Note that the training data for coefficient estiation and the testing data for error estiation are not overlapped. Hence, over-fitting can be easily detected. In addition, different groups should be selected for error estiation in different runs. As such, each run results in an error value e n (n =,,..., N that is easured fro a unique group of data points. The final odeling error is coputed as the average of {e n ; n =,,..., N}, i.e., e = (e + e e N /N. ore details on cross-validation can be found in [9]. 4. Suary Algorith suarizes the ajor steps of our proposed BF ethod. It consists of two core coponents: (i prior distribution definition, and (ii AP estiation. The efficacy of BF will be further deonstrated by our nuerical exaples in the next section. Algorith : Bayesian odel Fusion (BF. Start fro the early-stage perforance odel f E (x in (6.. Define the prior distribution for the late-stage odel coefficients {α L, ; =,,..., } by (-( and (5.. Collect late-stage siulation saples {(x (k, f (k L ; k =,,, }. 4. Solve the late-stage odel coefficients {α L, ; =,,..., } based on (4 where σ is deterined by cross-validation. 5. NUERICAL EXAPLES In this section, several circuit exaples designed in a coercial n COS process are used to deonstrate the efficacy of the proposed BF ethod. Our objective is to build post-layout perforance odels for these circuits. For testing and coparison purposes, two different perforance odeling techniques are ipleented: (i the traditional sparse regression ethod based on OP [8], and (ii the proposed BF ethod. Here, the OP algorith is chosen for coparison, since it is one of the state-of-the-art techniques in the literature. When ipleenting BF, we use the scheatic-level siulation data to define our prior knowledge for post-layout perforance odeling. In each exaple, two different data sets, referred to as the training set and the testing set respectively, are generated by rando sapling based on post-layout transistor-level siulation. The training set is used for coefficient fitting, including crossvalidation. The testing set contains independent rando saples that are used for odel validation. All nuerical experients are run on a.9ghz Linux server with 4GB eory. 5. Ring Oscillator Shown in Figure (a is the siplified circuit scheatic of a ring oscillator designed in a coercial n COS process. In this exaple, there are 777 independent rando variables in total to odel device-level process variations, including both inter-die variations and rando isatches. Our objective is to approxiate three post-layout perforance etrics, power, frequency and phase noise, as linear functions of these 777 rando variables. Figure (b-(d show how the perforance odeling error varies with the nuber of post-layout training saples. Note that for both OP and BF, the odeling error decays as the nuber of saples increases. However, given the sae nuber of postlayout training saples, BF is able to achieve substantially higher accuracy than OP, especially if only few saples are available. Table further copares the odeling error and cost for OP and BF. The total cost for perforance odeling consists of two ajor portions: (i siulation cost (i.e., the cost of running a transistor-level siulator to generate all post-layout saples in the training set, and (ii fitting cost (i.e., the cost of solving all unknown odel coefficients. As shown in Table, the total odeling cost is doinated by transistor-level siulation in this exaple. BF achieves 9 runtie speed-up over OP without surrendering any accuracy. odeling Error (%.5.5 (a OP (Traditional BF (Proposed Nuber of Post-Layout Saples (b. OP (Traditional BF (Proposed Nuber of Post-Layout Saples Nuber of Post-Layout Saples (c (d Figure. A ring oscillator designed in a coercial n COS process is used as an exaple for perforance odeling where BF requires significantly less post-layout saples than OP to achieve the sae accuracy: (a siplified circuit scheatic of the ring oscillator, (b odeling error for power, (c odeling error for frequency, and (d odeling error for phase noise. Table. Perforance odeling error and cost for ring oscillator OP (Traditional BF (Proposed # of post-layout training saples 9 odeling error for power.77%.7% odeling error for frequency.65%.54% odeling error for phase noise.%.% Siulation cost (Hour.58.4 Fitting cost (Second Total odeling cost (Hour.58.4 odeling Error (% odeling Error (% 5. SRA Read Path Figure (a shows the siplified circuit scheatic of an SRA read path designed in a coercial n COS process. In this exaple, there are 667 independent rando variables to odel device-level process variations. Read delay is our circuit perforance of interest, and it is approxiated as a linear function of all device-level rando variables. Figure (b and Table copare the odeling error and cost for OP and BF. Siilar to the ring oscillator exaple, two iportant observations can be ade here. First, BF requires substantially less post-layout training saples than OP, in order. OP (Traditional BF (Proposed 5

6 to achieve the sae odeling accuracy. In this exaple, BF reduces the nuber of required saples fro 4 to without surrendering any accuracy. Figure 4 further plots the histogras of odeling error for both OP and BF with 4 and postlayout training saples, respectively. Second, but ore iportantly, since the total odeling cost is doinated by transistor-level siulation, BF successfully reduces the total odeling cost by reducing the nuber of required post-layout training saples. Copared to OP, BF achieves 4 runtie speed-up, as shown in Table. Tiing logic Nuber of Post-Layout Saples (a (b Figure. A siplified SRA read path designed in a coercial n COS process is used as an exaple for perforance odeling where BF requires significantly less post-layout saples than OP to achieve the sae accuracy: (a siplified circuit scheatic of the SRA read path, and (b odeling error for read delay. Table. Perforance odeling error and cost for SRA OP (Traditional BF (Proposed # of post-layout training saples 4 odeling error for read delay.%.99% Siulation cost (Hour Fitting cost (Second.56. Total odeling cost (Hour Nuber of Saples 5 4 odeling Error (% Nuber of Saples 6. CONCLUSIONS In this paper, a novel BF algorith is proposed for efficient high-diensional perforance odeling of coplex AS circuits with consideration of large-scale process variations. BF borrows the early-stage (e.g., scheatic-level siulation data to learn a odel teplate that is statistically represented as a prior distribution. Next, the odel teplate encoding our prior knowledge is further calibrated by very few late-stage (e.g., postlayout siulation saples to accurately create a late-stage perforance odel. As such, the coputational cost of highdiensional perforance odeling can be substantially reduced, since only few transistor-level siulations are required at the late 4 OP (Traditional BF (Proposed odeling Error (% odeling Error (% (a (b Figure 4. Histogras of odeling error are estiated fro the testing set for read delay: (a odeling error of OP with 4 post-layout training saples, and (b odeling error of BF with post-layout training saples. 5 4 stage. As is deonstrated by our circuit exaples designed in a coercial n COS process, the proposed BF ethod achieves up to 9 runtie speedup copared to the traditional odeling technique. In our future work, we will further apply BF to several practical applications such as statistical analysis of large-scale AS systes. 7. ACNOWLEDGEENTS This work has been supported in part by National Science Foundation and Intel Corporation. 8. REFERENCES [] Seiconductor Industry Associate, International Technology Roadap for Seiconductors,. [] X. Li, J. Le and L. Pileggi, Statistical Perforance odeling and Optiization, Now Publishers, 7. [] Z. Feng and P. Li, Perforance-oriented statistical paraeter reduction of paraeterized systes via reduced rank regression, IEEE ICCAD, pp , 6. [4] A. Singhee and R. Rutenbar, Beyond low-order statistical response surfaces: latent variable regression for efficient, highly nonlinear fitting, IEEE DAC, pp. 56-6, 7. [5] A. itev,. arefat, D. a and J. Wang, Principle Hessian direction based paraeter reduction for interconnect networks with process variation, IEEE ICCAD, pp. 6-67, 7. [6] T. cconaghy and G. Gielen, Teplate-free sybolic perforance odeling of analog circuits via canonical-for functions and genetic prograing, IEEE Trans. on CAD, vol. 8, no. 8, pp. 6-75, Aug. 9. [7] T. cconaghy, High-diensional statistical odeling and analysis of custo integrated circuits, IEEE CICC,. [8] X. Li, Finding deterinistic solution fro underdeterined equation: large-scale perforance odeling of analog/rf circuits, IEEE Trans. on CAD, vol. 9, no., pp , Nov.. [9] X. Li, J. Le, P. Gopalakrishnan and L. Pileggi, Asyptotic probability extraction for nonnoral perforance distributions, IEEE Trans. on CAD, vol. 6, no., pp. 6-7, Jan. 7. []. Sengupta, S. Saxena, L. Daldoss, G. raer, S. inehane and J. Cheng, Application-specific worst case corners using response surfaces and statistical odels, IEEE Trans. on CAD, vol. 4, no. 9, pp. 7-8, 5. [] Z. Wang and S. Director, An efficient yield optiization ethod using a two step linear approxiation of circuit perforance, IEEE EDAC, pp , 994. [] A. Dharchoudhury and S. ang, Worse-case analysis and optiization of VLSI circuit perforance, IEEE Trans. on CAD, vol. 4, no. 4, pp , Apr [] G. Debyser and G. Gielen, Efficient analog circuit synthesis with siultaneous yield and robustness optiization, IEEE ICCAD, pp. 8-, 998. [4] F. Schenkel,. Pronath, S. Zizala, R. Schwencker, H. Graeb and. Antreich, isatch analysis and direct yield optiization by spec-wise linearization and feasibility-guided search, IEEE DAC, pp ,. [5] X. Li, P. Gopalakrishnan, Y. Xu and L. Pileggi, Robust analog/rf circuit design with projection-based perforance odeling, IEEE Trans. on CAD, vol. 6, no., pp. -5, Jan. 7. [6] X. Li, W. Zhang, F. Wang, S. Sun and C. Gu, Efficient paraetric yield estiation of analog/ixed-signal circuits via Bayesian odel fusion, IEEE ICCAD,. [7] S. Ji, Y. Xue, and L. Carin, Bayesian copressive sensing, IEEE Trans. on Signal Processing, pp , Jun. 8. [8] R. yers and D. ontgoery, Response Surface ethodology: Process and Product Optiization Using Designed Experients, Wiley-Interscience,. [9] C. Bishop, Pattern Recognition and achine Learning, Springer, 6. 6

This paper studies a rental firm that offers reusable products to price- and quality-of-service sensitive

This paper studies a rental firm that offers reusable products to price- and quality-of-service sensitive MANUFACTURING & SERVICE OPERATIONS MANAGEMENT Vol., No. 3, Suer 28, pp. 429 447 issn 523-464 eissn 526-5498 8 3 429 infors doi.287/so.7.8 28 INFORMS INFORMS holds copyright to this article and distributed

More information

Machine Learning Applications in Grid Computing

Machine Learning Applications in Grid Computing Machine Learning Applications in Grid Coputing George Cybenko, Guofei Jiang and Daniel Bilar Thayer School of Engineering Dartouth College Hanover, NH 03755, USA gvc@dartouth.edu, guofei.jiang@dartouth.edu

More information

Design of Model Reference Self Tuning Mechanism for PID like Fuzzy Controller

Design of Model Reference Self Tuning Mechanism for PID like Fuzzy Controller Research Article International Journal of Current Engineering and Technology EISSN 77 46, PISSN 347 56 4 INPRESSCO, All Rights Reserved Available at http://inpressco.co/category/ijcet Design of Model Reference

More information

Modeling operational risk data reported above a time-varying threshold

Modeling operational risk data reported above a time-varying threshold Modeling operational risk data reported above a tie-varying threshold Pavel V. Shevchenko CSIRO Matheatical and Inforation Sciences, Sydney, Locked bag 7, North Ryde, NSW, 670, Australia. e-ail: Pavel.Shevchenko@csiro.au

More information

Extended-Horizon Analysis of Pressure Sensitivities for Leak Detection in Water Distribution Networks: Application to the Barcelona Network

Extended-Horizon Analysis of Pressure Sensitivities for Leak Detection in Water Distribution Networks: Application to the Barcelona Network 2013 European Control Conference (ECC) July 17-19, 2013, Zürich, Switzerland. Extended-Horizon Analysis of Pressure Sensitivities for Leak Detection in Water Distribution Networks: Application to the Barcelona

More information

Reliability Constrained Packet-sizing for Linear Multi-hop Wireless Networks

Reliability Constrained Packet-sizing for Linear Multi-hop Wireless Networks Reliability Constrained acket-sizing for inear Multi-hop Wireless Networks Ning Wen, and Randall A. Berry Departent of Electrical Engineering and Coputer Science Northwestern University, Evanston, Illinois

More information

Use of extrapolation to forecast the working capital in the mechanical engineering companies

Use of extrapolation to forecast the working capital in the mechanical engineering companies ECONTECHMOD. AN INTERNATIONAL QUARTERLY JOURNAL 2014. Vol. 1. No. 1. 23 28 Use of extrapolation to forecast the working capital in the echanical engineering copanies A. Cherep, Y. Shvets Departent of finance

More information

A CHAOS MODEL OF SUBHARMONIC OSCILLATIONS IN CURRENT MODE PWM BOOST CONVERTERS

A CHAOS MODEL OF SUBHARMONIC OSCILLATIONS IN CURRENT MODE PWM BOOST CONVERTERS A CHAOS MODEL OF SUBHARMONIC OSCILLATIONS IN CURRENT MODE PWM BOOST CONVERTERS Isaac Zafrany and Sa BenYaakov Departent of Electrical and Coputer Engineering BenGurion University of the Negev P. O. Box

More information

Searching strategy for multi-target discovery in wireless networks

Searching strategy for multi-target discovery in wireless networks Searching strategy for ulti-target discovery in wireless networks Zhao Cheng, Wendi B. Heinzelan Departent of Electrical and Coputer Engineering University of Rochester Rochester, NY 467 (585) 75-{878,

More information

Image restoration for a rectangular poor-pixels detector

Image restoration for a rectangular poor-pixels detector Iage restoration for a rectangular poor-pixels detector Pengcheng Wen 1, Xiangjun Wang 1, Hong Wei 2 1 State Key Laboratory of Precision Measuring Technology and Instruents, Tianjin University, China 2

More information

Equivalent Tapped Delay Line Channel Responses with Reduced Taps

Equivalent Tapped Delay Line Channel Responses with Reduced Taps Equivalent Tapped Delay Line Channel Responses with Reduced Taps Shweta Sagari, Wade Trappe, Larry Greenstein {shsagari, trappe, ljg}@winlab.rutgers.edu WINLAB, Rutgers University, North Brunswick, NJ

More information

CRM FACTORS ASSESSMENT USING ANALYTIC HIERARCHY PROCESS

CRM FACTORS ASSESSMENT USING ANALYTIC HIERARCHY PROCESS 641 CRM FACTORS ASSESSMENT USING ANALYTIC HIERARCHY PROCESS Marketa Zajarosova 1* *Ph.D. VSB - Technical University of Ostrava, THE CZECH REPUBLIC arketa.zajarosova@vsb.cz Abstract Custoer relationship

More information

6. Time (or Space) Series Analysis

6. Time (or Space) Series Analysis ATM 55 otes: Tie Series Analysis - Section 6a Page 8 6. Tie (or Space) Series Analysis In this chapter we will consider soe coon aspects of tie series analysis including autocorrelation, statistical prediction,

More information

Analyzing Spatiotemporal Characteristics of Education Network Traffic with Flexible Multiscale Entropy

Analyzing Spatiotemporal Characteristics of Education Network Traffic with Flexible Multiscale Entropy Vol. 9, No. 5 (2016), pp.303-312 http://dx.doi.org/10.14257/ijgdc.2016.9.5.26 Analyzing Spatioteporal Characteristics of Education Network Traffic with Flexible Multiscale Entropy Chen Yang, Renjie Zhou

More information

Markovian inventory policy with application to the paper industry

Markovian inventory policy with application to the paper industry Coputers and Cheical Engineering 26 (2002) 1399 1413 www.elsevier.co/locate/copcheeng Markovian inventory policy with application to the paper industry K. Karen Yin a, *, Hu Liu a,1, Neil E. Johnson b,2

More information

International Journal of Management & Information Systems First Quarter 2012 Volume 16, Number 1

International Journal of Management & Information Systems First Quarter 2012 Volume 16, Number 1 International Journal of Manageent & Inforation Systes First Quarter 2012 Volue 16, Nuber 1 Proposal And Effectiveness Of A Highly Copelling Direct Mail Method - Establishent And Deployent Of PMOS-DM Hisatoshi

More information

RECURSIVE DYNAMIC PROGRAMMING: HEURISTIC RULES, BOUNDING AND STATE SPACE REDUCTION. Henrik Kure

RECURSIVE DYNAMIC PROGRAMMING: HEURISTIC RULES, BOUNDING AND STATE SPACE REDUCTION. Henrik Kure RECURSIVE DYNAMIC PROGRAMMING: HEURISTIC RULES, BOUNDING AND STATE SPACE REDUCTION Henrik Kure Dina, Danish Inforatics Network In the Agricultural Sciences Royal Veterinary and Agricultural University

More information

Calculating the Return on Investment (ROI) for DMSMS Management. The Problem with Cost Avoidance

Calculating the Return on Investment (ROI) for DMSMS Management. The Problem with Cost Avoidance Calculating the Return on nvestent () for DMSMS Manageent Peter Sandborn CALCE, Departent of Mechanical Engineering (31) 45-3167 sandborn@calce.ud.edu www.ene.ud.edu/escml/obsolescence.ht October 28, 21

More information

An Innovate Dynamic Load Balancing Algorithm Based on Task

An Innovate Dynamic Load Balancing Algorithm Based on Task An Innovate Dynaic Load Balancing Algorith Based on Task Classification Hong-bin Wang,,a, Zhi-yi Fang, b, Guan-nan Qu,*,c, Xiao-dan Ren,d College of Coputer Science and Technology, Jilin University, Changchun

More information

Adaptive Modulation and Coding for Unmanned Aerial Vehicle (UAV) Radio Channel

Adaptive Modulation and Coding for Unmanned Aerial Vehicle (UAV) Radio Channel Recent Advances in Counications Adaptive odulation and Coding for Unanned Aerial Vehicle (UAV) Radio Channel Airhossein Fereidountabar,Gian Carlo Cardarilli, Rocco Fazzolari,Luca Di Nunzio Abstract In

More information

Factor Model. Arbitrage Pricing Theory. Systematic Versus Non-Systematic Risk. Intuitive Argument

Factor Model. Arbitrage Pricing Theory. Systematic Versus Non-Systematic Risk. Intuitive Argument Ross [1],[]) presents the aritrage pricing theory. The idea is that the structure of asset returns leads naturally to a odel of risk preia, for otherwise there would exist an opportunity for aritrage profit.

More information

ASIC Design Project Management Supported by Multi Agent Simulation

ASIC Design Project Management Supported by Multi Agent Simulation ASIC Design Project Manageent Supported by Multi Agent Siulation Jana Blaschke, Christian Sebeke, Wolfgang Rosenstiel Abstract The coplexity of Application Specific Integrated Circuits (ASICs) is continuously

More information

Media Adaptation Framework in Biofeedback System for Stroke Patient Rehabilitation

Media Adaptation Framework in Biofeedback System for Stroke Patient Rehabilitation Media Adaptation Fraework in Biofeedback Syste for Stroke Patient Rehabilitation Yinpeng Chen, Weiwei Xu, Hari Sundara, Thanassis Rikakis, Sheng-Min Liu Arts, Media and Engineering Progra Arizona State

More information

Resource Allocation in Wireless Networks with Multiple Relays

Resource Allocation in Wireless Networks with Multiple Relays Resource Allocation in Wireless Networks with Multiple Relays Kağan Bakanoğlu, Stefano Toasin, Elza Erkip Departent of Electrical and Coputer Engineering, Polytechnic Institute of NYU, Brooklyn, NY, 0

More information

arxiv:0805.1434v1 [math.pr] 9 May 2008

arxiv:0805.1434v1 [math.pr] 9 May 2008 Degree-distribution stability of scale-free networs Zhenting Hou, Xiangxing Kong, Dinghua Shi,2, and Guanrong Chen 3 School of Matheatics, Central South University, Changsha 40083, China 2 Departent of

More information

CLOSED-LOOP SUPPLY CHAIN NETWORK OPTIMIZATION FOR HONG KONG CARTRIDGE RECYCLING INDUSTRY

CLOSED-LOOP SUPPLY CHAIN NETWORK OPTIMIZATION FOR HONG KONG CARTRIDGE RECYCLING INDUSTRY CLOSED-LOOP SUPPLY CHAIN NETWORK OPTIMIZATION FOR HONG KONG CARTRIDGE RECYCLING INDUSTRY Y. T. Chen Departent of Industrial and Systes Engineering Hong Kong Polytechnic University, Hong Kong yongtong.chen@connect.polyu.hk

More information

Evaluating Inventory Management Performance: a Preliminary Desk-Simulation Study Based on IOC Model

Evaluating Inventory Management Performance: a Preliminary Desk-Simulation Study Based on IOC Model Evaluating Inventory Manageent Perforance: a Preliinary Desk-Siulation Study Based on IOC Model Flora Bernardel, Roberto Panizzolo, and Davide Martinazzo Abstract The focus of this study is on preliinary

More information

AN ALGORITHM FOR REDUCING THE DIMENSION AND SIZE OF A SAMPLE FOR DATA EXPLORATION PROCEDURES

AN ALGORITHM FOR REDUCING THE DIMENSION AND SIZE OF A SAMPLE FOR DATA EXPLORATION PROCEDURES Int. J. Appl. Math. Coput. Sci., 2014, Vol. 24, No. 1, 133 149 DOI: 10.2478/acs-2014-0011 AN ALGORITHM FOR REDUCING THE DIMENSION AND SIZE OF A SAMPLE FOR DATA EXPLORATION PROCEDURES PIOTR KULCZYCKI,,

More information

Dynamic Placement for Clustered Web Applications

Dynamic Placement for Clustered Web Applications Dynaic laceent for Clustered Web Applications A. Karve, T. Kibrel, G. acifici, M. Spreitzer, M. Steinder, M. Sviridenko, and A. Tantawi IBM T.J. Watson Research Center {karve,kibrel,giovanni,spreitz,steinder,sviri,tantawi}@us.ib.co

More information

A Scalable Application Placement Controller for Enterprise Data Centers

A Scalable Application Placement Controller for Enterprise Data Centers W WWW 7 / Track: Perforance and Scalability A Scalable Application Placeent Controller for Enterprise Data Centers Chunqiang Tang, Malgorzata Steinder, Michael Spreitzer, and Giovanni Pacifici IBM T.J.

More information

Method of supply chain optimization in E-commerce

Method of supply chain optimization in E-commerce MPRA Munich Personal RePEc Archive Method of supply chain optiization in E-coerce Petr Suchánek and Robert Bucki Silesian University - School of Business Adinistration, The College of Inforatics and Manageent

More information

PERFORMANCE METRICS FOR THE IT SERVICES PORTFOLIO

PERFORMANCE METRICS FOR THE IT SERVICES PORTFOLIO Bulletin of the Transilvania University of Braşov Series I: Engineering Sciences Vol. 4 (53) No. - 0 PERFORMANCE METRICS FOR THE IT SERVICES PORTFOLIO V. CAZACU I. SZÉKELY F. SANDU 3 T. BĂLAN Abstract:

More information

The Virtual Spring Mass System

The Virtual Spring Mass System The Virtual Spring Mass Syste J. S. Freudenberg EECS 6 Ebedded Control Systes Huan Coputer Interaction A force feedbac syste, such as the haptic heel used in the EECS 6 lab, is capable of exhibiting a

More information

Optimal Resource-Constraint Project Scheduling with Overlapping Modes

Optimal Resource-Constraint Project Scheduling with Overlapping Modes Optial Resource-Constraint Proect Scheduling with Overlapping Modes François Berthaut Lucas Grèze Robert Pellerin Nathalie Perrier Adnène Hai February 20 CIRRELT-20-09 Bureaux de Montréal : Bureaux de

More information

Data Set Generation for Rectangular Placement Problems

Data Set Generation for Rectangular Placement Problems Data Set Generation for Rectangular Placeent Probles Christine L. Valenzuela (Muford) Pearl Y. Wang School of Coputer Science & Inforatics Departent of Coputer Science MS 4A5 Cardiff University George

More information

The individual neurons are complicated. They have a myriad of parts, subsystems and control mechanisms. They convey information via a host of

The individual neurons are complicated. They have a myriad of parts, subsystems and control mechanisms. They convey information via a host of CHAPTER 4 ARTIFICIAL NEURAL NETWORKS 4. INTRODUCTION Artificial Neural Networks (ANNs) are relatively crude electronic odels based on the neural structure of the brain. The brain learns fro experience.

More information

Managing Complex Network Operation with Predictive Analytics

Managing Complex Network Operation with Predictive Analytics Managing Coplex Network Operation with Predictive Analytics Zhenyu Huang, Pak Chung Wong, Patrick Mackey, Yousu Chen, Jian Ma, Kevin Schneider, and Frank L. Greitzer Pacific Northwest National Laboratory

More information

Online Bagging and Boosting

Online Bagging and Boosting Abstract Bagging and boosting are two of the ost well-known enseble learning ethods due to their theoretical perforance guarantees and strong experiental results. However, these algoriths have been used

More information

Modified Latin Hypercube Sampling Monte Carlo (MLHSMC) Estimation for Average Quality Index

Modified Latin Hypercube Sampling Monte Carlo (MLHSMC) Estimation for Average Quality Index Analog Integrated Circuits and Signal Processing, vol. 9, no., April 999. Abstract Modified Latin Hypercube Sapling Monte Carlo (MLHSMC) Estiation for Average Quality Index Mansour Keraat and Richard Kielbasa

More information

Cooperative Caching for Adaptive Bit Rate Streaming in Content Delivery Networks

Cooperative Caching for Adaptive Bit Rate Streaming in Content Delivery Networks Cooperative Caching for Adaptive Bit Rate Streaing in Content Delivery Networs Phuong Luu Vo Departent of Coputer Science and Engineering, International University - VNUHCM, Vietna vtlphuong@hciu.edu.vn

More information

Airline Yield Management with Overbooking, Cancellations, and No-Shows JANAKIRAM SUBRAMANIAN

Airline Yield Management with Overbooking, Cancellations, and No-Shows JANAKIRAM SUBRAMANIAN Airline Yield Manageent with Overbooking, Cancellations, and No-Shows JANAKIRAM SUBRAMANIAN Integral Developent Corporation, 301 University Avenue, Suite 200, Palo Alto, California 94301 SHALER STIDHAM

More information

CPU Animation. Introduction. CPU skinning. CPUSkin Scalar:

CPU Animation. Introduction. CPU skinning. CPUSkin Scalar: CPU Aniation Introduction The iportance of real-tie character aniation has greatly increased in odern gaes. Aniating eshes ia 'skinning' can be perfored on both a general purpose CPU and a ore specialized

More information

Modeling Cooperative Gene Regulation Using Fast Orthogonal Search

Modeling Cooperative Gene Regulation Using Fast Orthogonal Search 8 The Open Bioinforatics Journal, 28, 2, 8-89 Open Access odeling Cooperative Gene Regulation Using Fast Orthogonal Search Ian inz* and ichael J. Korenberg* Departent of Electrical and Coputer Engineering,

More information

Implementation of Active Queue Management in a Combined Input and Output Queued Switch

Implementation of Active Queue Management in a Combined Input and Output Queued Switch pleentation of Active Queue Manageent in a obined nput and Output Queued Switch Bartek Wydrowski and Moshe Zukeran AR Special Research entre for Ultra-Broadband nforation Networks, EEE Departent, The University

More information

A quantum secret ballot. Abstract

A quantum secret ballot. Abstract A quantu secret ballot Shahar Dolev and Itaar Pitowsky The Edelstein Center, Levi Building, The Hebrerw University, Givat Ra, Jerusale, Israel Boaz Tair arxiv:quant-ph/060087v 8 Mar 006 Departent of Philosophy

More information

Energy Efficient VM Scheduling for Cloud Data Centers: Exact allocation and migration algorithms

Energy Efficient VM Scheduling for Cloud Data Centers: Exact allocation and migration algorithms Energy Efficient VM Scheduling for Cloud Data Centers: Exact allocation and igration algoriths Chaia Ghribi, Makhlouf Hadji and Djaal Zeghlache Institut Mines-Téléco, Téléco SudParis UMR CNRS 5157 9, Rue

More information

Pure Bending Determination of Stress-Strain Curves for an Aluminum Alloy

Pure Bending Determination of Stress-Strain Curves for an Aluminum Alloy Proceedings of the World Congress on Engineering 0 Vol III WCE 0, July 6-8, 0, London, U.K. Pure Bending Deterination of Stress-Strain Curves for an Aluinu Alloy D. Torres-Franco, G. Urriolagoitia-Sosa,

More information

An Optimal Task Allocation Model for System Cost Analysis in Heterogeneous Distributed Computing Systems: A Heuristic Approach

An Optimal Task Allocation Model for System Cost Analysis in Heterogeneous Distributed Computing Systems: A Heuristic Approach An Optial Tas Allocation Model for Syste Cost Analysis in Heterogeneous Distributed Coputing Systes: A Heuristic Approach P. K. Yadav Central Building Research Institute, Rooree- 247667, Uttarahand (INDIA)

More information

High Performance Chinese/English Mixed OCR with Character Level Language Identification

High Performance Chinese/English Mixed OCR with Character Level Language Identification 2009 0th International Conference on Docuent Analysis and Recognition High Perforance Chinese/English Mixed OCR with Character Level Language Identification Kai Wang Institute of Machine Intelligence,

More information

Applying Multiple Neural Networks on Large Scale Data

Applying Multiple Neural Networks on Large Scale Data 0 International Conference on Inforation and Electronics Engineering IPCSIT vol6 (0) (0) IACSIT Press, Singapore Applying Multiple Neural Networks on Large Scale Data Kritsanatt Boonkiatpong and Sukree

More information

Physics 211: Lab Oscillations. Simple Harmonic Motion.

Physics 211: Lab Oscillations. Simple Harmonic Motion. Physics 11: Lab Oscillations. Siple Haronic Motion. Reading Assignent: Chapter 15 Introduction: As we learned in class, physical systes will undergo an oscillatory otion, when displaced fro a stable equilibriu.

More information

Audio Engineering Society. Convention Paper. Presented at the 119th Convention 2005 October 7 10 New York, New York USA

Audio Engineering Society. Convention Paper. Presented at the 119th Convention 2005 October 7 10 New York, New York USA Audio Engineering Society Convention Paper Presented at the 119th Convention 2005 October 7 10 New York, New York USA This convention paper has been reproduced fro the authors advance anuscript, without

More information

Exploiting Hardware Heterogeneity within the Same Instance Type of Amazon EC2

Exploiting Hardware Heterogeneity within the Same Instance Type of Amazon EC2 Exploiting Hardware Heterogeneity within the Sae Instance Type of Aazon EC2 Zhonghong Ou, Hao Zhuang, Jukka K. Nurinen, Antti Ylä-Jääski, Pan Hui Aalto University, Finland; Deutsch Teleko Laboratories,

More information

Fuzzy Sets in HR Management

Fuzzy Sets in HR Management Acta Polytechnica Hungarica Vol. 8, No. 3, 2011 Fuzzy Sets in HR Manageent Blanka Zeková AXIOM SW, s.r.o., 760 01 Zlín, Czech Republic blanka.zekova@sezna.cz Jana Talašová Faculty of Science, Palacký Univerzity,

More information

SOME APPLICATIONS OF FORECASTING Prof. Thomas B. Fomby Department of Economics Southern Methodist University May 2008

SOME APPLICATIONS OF FORECASTING Prof. Thomas B. Fomby Department of Economics Southern Methodist University May 2008 SOME APPLCATONS OF FORECASTNG Prof. Thoas B. Foby Departent of Econoics Southern Methodist University May 8 To deonstrate the usefulness of forecasting ethods this note discusses four applications of forecasting

More information

Introduction to Unit Conversion: the SI

Introduction to Unit Conversion: the SI The Matheatics 11 Copetency Test Introduction to Unit Conversion: the SI In this the next docuent in this series is presented illustrated an effective reliable approach to carryin out unit conversions

More information

The Benefit of SMT in the Multi-Core Era: Flexibility towards Degrees of Thread-Level Parallelism

The Benefit of SMT in the Multi-Core Era: Flexibility towards Degrees of Thread-Level Parallelism The enefit of SMT in the Multi-Core Era: Flexibility towards Degrees of Thread-Level Parallelis Stijn Eyeran Lieven Eeckhout Ghent University, elgiu Stijn.Eyeran@elis.UGent.be, Lieven.Eeckhout@elis.UGent.be

More information

Partitioned Elias-Fano Indexes

Partitioned Elias-Fano Indexes Partitioned Elias-ano Indexes Giuseppe Ottaviano ISTI-CNR, Pisa giuseppe.ottaviano@isti.cnr.it Rossano Venturini Dept. of Coputer Science, University of Pisa rossano@di.unipi.it ABSTRACT The Elias-ano

More information

Comment on On Discriminative vs. Generative Classifiers: A Comparison of Logistic Regression and Naive Bayes

Comment on On Discriminative vs. Generative Classifiers: A Comparison of Logistic Regression and Naive Bayes Coent on On Discriinative vs. Generative Classifiers: A Coparison of Logistic Regression and Naive Bayes Jing-Hao Xue (jinghao@stats.gla.ac.uk) and D. Michael Titterington (ike@stats.gla.ac.uk) Departent

More information

Modeling Parallel Applications Performance on Heterogeneous Systems

Modeling Parallel Applications Performance on Heterogeneous Systems Modeling Parallel Applications Perforance on Heterogeneous Systes Jaeela Al-Jaroodi, Nader Mohaed, Hong Jiang and David Swanson Departent of Coputer Science and Engineering University of Nebraska Lincoln

More information

Evaluating the Effectiveness of Task Overlapping as a Risk Response Strategy in Engineering Projects

Evaluating the Effectiveness of Task Overlapping as a Risk Response Strategy in Engineering Projects Evaluating the Effectiveness of Task Overlapping as a Risk Response Strategy in Engineering Projects Lucas Grèze Robert Pellerin Nathalie Perrier Patrice Leclaire February 2011 CIRRELT-2011-11 Bureaux

More information

The Fundamentals of Modal Testing

The Fundamentals of Modal Testing The Fundaentals of Modal Testing Application Note 243-3 Η(ω) = Σ n r=1 φ φ i j / 2 2 2 2 ( ω n - ω ) + (2ξωωn) Preface Modal analysis is defined as the study of the dynaic characteristics of a echanical

More information

Data Streaming Algorithms for Estimating Entropy of Network Traffic

Data Streaming Algorithms for Estimating Entropy of Network Traffic Data Streaing Algoriths for Estiating Entropy of Network Traffic Ashwin Lall University of Rochester Vyas Sekar Carnegie Mellon University Mitsunori Ogihara University of Rochester Jun (Ji) Xu Georgia

More information

Capacity of Multiple-Antenna Systems With Both Receiver and Transmitter Channel State Information

Capacity of Multiple-Antenna Systems With Both Receiver and Transmitter Channel State Information IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 49, NO., OCTOBER 23 2697 Capacity of Multiple-Antenna Systes With Both Receiver and Transitter Channel State Inforation Sudharan K. Jayaweera, Student Meber,

More information

Halloween Costume Ideas for the Wii Game

Halloween Costume Ideas for the Wii Game Algorithica 2001) 30: 101 139 DOI: 101007/s00453-001-0003-0 Algorithica 2001 Springer-Verlag New York Inc Optial Search and One-Way Trading Online Algoriths R El-Yaniv, 1 A Fiat, 2 R M Karp, 3 and G Turpin

More information

On Computing Nearest Neighbors with Applications to Decoding of Binary Linear Codes

On Computing Nearest Neighbors with Applications to Decoding of Binary Linear Codes On Coputing Nearest Neighbors with Applications to Decoding of Binary Linear Codes Alexander May and Ilya Ozerov Horst Görtz Institute for IT-Security Ruhr-University Bochu, Gerany Faculty of Matheatics

More information

An Integrated Approach for Monitoring Service Level Parameters of Software-Defined Networking

An Integrated Approach for Monitoring Service Level Parameters of Software-Defined Networking International Journal of Future Generation Counication and Networking Vol. 8, No. 6 (15), pp. 197-4 http://d.doi.org/1.1457/ijfgcn.15.8.6.19 An Integrated Approach for Monitoring Service Level Paraeters

More information

Trading Regret for Efficiency: Online Convex Optimization with Long Term Constraints

Trading Regret for Efficiency: Online Convex Optimization with Long Term Constraints Journal of Machine Learning Research 13 2012) 2503-2528 Subitted 8/11; Revised 3/12; Published 9/12 rading Regret for Efficiency: Online Convex Optiization with Long er Constraints Mehrdad Mahdavi Rong

More information

Gaussian Processes for Regression: A Quick Introduction

Gaussian Processes for Regression: A Quick Introduction Gaussian Processes for Regression A Quick Introduction M Ebden, August 28 Coents to arkebden@engoacuk MOTIVATION Figure illustrates a typical eaple of a prediction proble given soe noisy observations of

More information

An Improved Decision-making Model of Human Resource Outsourcing Based on Internet Collaboration

An Improved Decision-making Model of Human Resource Outsourcing Based on Internet Collaboration International Journal of Hybrid Inforation Technology, pp. 339-350 http://dx.doi.org/10.14257/hit.2016.9.4.28 An Iproved Decision-aking Model of Huan Resource Outsourcing Based on Internet Collaboration

More information

SAMPLING METHODS LEARNING OBJECTIVES

SAMPLING METHODS LEARNING OBJECTIVES 6 SAMPLING METHODS 6 Using Statistics 6-6 2 Nonprobability Sapling and Bias 6-6 Stratified Rando Sapling 6-2 6 4 Cluster Sapling 6-4 6 5 Systeatic Sapling 6-9 6 6 Nonresponse 6-2 6 7 Suary and Review of

More information

The Velocities of Gas Molecules

The Velocities of Gas Molecules he Velocities of Gas Molecules by Flick Colean Departent of Cheistry Wellesley College Wellesley MA 8 Copyright Flick Colean 996 All rights reserved You are welcoe to use this docuent in your own classes

More information

Budget-optimal Crowdsourcing using Low-rank Matrix Approximations

Budget-optimal Crowdsourcing using Low-rank Matrix Approximations Budget-optial Crowdsourcing using Low-rank Matrix Approxiations David R. Karger, Sewoong Oh, and Devavrat Shah Departent of EECS, Massachusetts Institute of Technology Eail: {karger, swoh, devavrat}@it.edu

More information

Preference-based Search and Multi-criteria Optimization

Preference-based Search and Multi-criteria Optimization Fro: AAAI-02 Proceedings. Copyright 2002, AAAI (www.aaai.org). All rights reserved. Preference-based Search and Multi-criteria Optiization Ulrich Junker ILOG 1681, route des Dolines F-06560 Valbonne ujunker@ilog.fr

More information

An Approach to Combating Free-riding in Peer-to-Peer Networks

An Approach to Combating Free-riding in Peer-to-Peer Networks An Approach to Cobating Free-riding in Peer-to-Peer Networks Victor Ponce, Jie Wu, and Xiuqi Li Departent of Coputer Science and Engineering Florida Atlantic University Boca Raton, FL 33431 April 7, 2008

More information

REQUIREMENTS FOR A COMPUTER SCIENCE CURRICULUM EMPHASIZING INFORMATION TECHNOLOGY SUBJECT AREA: CURRICULUM ISSUES

REQUIREMENTS FOR A COMPUTER SCIENCE CURRICULUM EMPHASIZING INFORMATION TECHNOLOGY SUBJECT AREA: CURRICULUM ISSUES REQUIREMENTS FOR A COMPUTER SCIENCE CURRICULUM EMPHASIZING INFORMATION TECHNOLOGY SUBJECT AREA: CURRICULUM ISSUES Charles Reynolds Christopher Fox reynolds @cs.ju.edu fox@cs.ju.edu Departent of Coputer

More information

Information Processing Letters

Information Processing Letters Inforation Processing Letters 111 2011) 178 183 Contents lists available at ScienceDirect Inforation Processing Letters www.elsevier.co/locate/ipl Offline file assignents for online load balancing Paul

More information

Lecture L26-3D Rigid Body Dynamics: The Inertia Tensor

Lecture L26-3D Rigid Body Dynamics: The Inertia Tensor J. Peraire, S. Widnall 16.07 Dynaics Fall 008 Lecture L6-3D Rigid Body Dynaics: The Inertia Tensor Version.1 In this lecture, we will derive an expression for the angular oentu of a 3D rigid body. We shall

More information

Real Time Target Tracking with Binary Sensor Networks and Parallel Computing

Real Time Target Tracking with Binary Sensor Networks and Parallel Computing Real Tie Target Tracking with Binary Sensor Networks and Parallel Coputing Hong Lin, John Rushing, Sara J. Graves, Steve Tanner, and Evans Criswell Abstract A parallel real tie data fusion and target tracking

More information

COMBINING CRASH RECORDER AND PAIRED COMPARISON TECHNIQUE: INJURY RISK FUNCTIONS IN FRONTAL AND REAR IMPACTS WITH SPECIAL REFERENCE TO NECK INJURIES

COMBINING CRASH RECORDER AND PAIRED COMPARISON TECHNIQUE: INJURY RISK FUNCTIONS IN FRONTAL AND REAR IMPACTS WITH SPECIAL REFERENCE TO NECK INJURIES COMBINING CRASH RECORDER AND AIRED COMARISON TECHNIQUE: INJURY RISK FUNCTIONS IN FRONTAL AND REAR IMACTS WITH SECIAL REFERENCE TO NECK INJURIES Anders Kullgren, Maria Krafft Folksa Research, 66 Stockhol,

More information

Construction Economics & Finance. Module 3 Lecture-1

Construction Economics & Finance. Module 3 Lecture-1 Depreciation:- Construction Econoics & Finance Module 3 Lecture- It represents the reduction in arket value of an asset due to age, wear and tear and obsolescence. The physical deterioration of the asset

More information

Quality evaluation of the model-based forecasts of implied volatility index

Quality evaluation of the model-based forecasts of implied volatility index Quality evaluation of the odel-based forecasts of iplied volatility index Katarzyna Łęczycka 1 Abstract Influence of volatility on financial arket forecasts is very high. It appears as a specific factor

More information

Position Auctions and Non-uniform Conversion Rates

Position Auctions and Non-uniform Conversion Rates Position Auctions and Non-unifor Conversion Rates Liad Blurosen Microsoft Research Mountain View, CA 944 liadbl@icrosoft.co Jason D. Hartline Shuzhen Nong Electrical Engineering and Microsoft AdCenter

More information

Evaluating Software Quality of Vendors using Fuzzy Analytic Hierarchy Process

Evaluating Software Quality of Vendors using Fuzzy Analytic Hierarchy Process IMECS 2008 9-2 March 2008 Hong Kong Evaluating Software Quality of Vendors using Fuzzy Analytic Hierarchy Process Kevin K.F. Yuen* Henry C.W. au Abstract This paper proposes a fuzzy Analytic Hierarchy

More information

IT SOURCING PORTFOLIO MANAGEMENT FOR IT SERVICES PROVIDERS - A RISK/COST PERSPECTIVE

IT SOURCING PORTFOLIO MANAGEMENT FOR IT SERVICES PROVIDERS - A RISK/COST PERSPECTIVE IT SOURCING PORTFOLIO MANAGEMENT FOR IT SERVICES PROVIDERS - A RISK/COST PERSPECTIVE Copleted Research Paper Steffen Zierann Arne Katzarzik Dennis Kundisch Abstract Utilizing a global IT sourcing strategy

More information

ESTIMATING LIQUIDITY PREMIA IN THE SPANISH GOVERNMENT SECURITIES MARKET

ESTIMATING LIQUIDITY PREMIA IN THE SPANISH GOVERNMENT SECURITIES MARKET ESTIMATING LIQUIDITY PREMIA IN THE SPANISH GOVERNMENT SECURITIES MARKET Francisco Alonso, Roberto Blanco, Ana del Río and Alicia Sanchis Banco de España Banco de España Servicio de Estudios Docuento de

More information

Calculation Method for evaluating Solar Assisted Heat Pump Systems in SAP 2009. 15 July 2013

Calculation Method for evaluating Solar Assisted Heat Pump Systems in SAP 2009. 15 July 2013 Calculation Method for evaluating Solar Assisted Heat Pup Systes in SAP 2009 15 July 2013 Page 1 of 17 1 Introduction This docuent describes how Solar Assisted Heat Pup Systes are recognised in the National

More information

A Fast Algorithm for Online Placement and Reorganization of Replicated Data

A Fast Algorithm for Online Placement and Reorganization of Replicated Data A Fast Algorith for Online Placeent and Reorganization of Replicated Data R. J. Honicky Storage Systes Research Center University of California, Santa Cruz Ethan L. Miller Storage Systes Research Center

More information

ABSTRACT KEYWORDS. Comonotonicity, dependence, correlation, concordance, copula, multivariate. 1. INTRODUCTION

ABSTRACT KEYWORDS. Comonotonicity, dependence, correlation, concordance, copula, multivariate. 1. INTRODUCTION MEASURING COMONOTONICITY IN M-DIMENSIONAL VECTORS BY INGE KOCH AND ANN DE SCHEPPER ABSTRACT In this contribution, a new easure of coonotonicity for -diensional vectors is introduced, with values between

More information

INTEGRATED ENVIRONMENT FOR STORING AND HANDLING INFORMATION IN TASKS OF INDUCTIVE MODELLING FOR BUSINESS INTELLIGENCE SYSTEMS

INTEGRATED ENVIRONMENT FOR STORING AND HANDLING INFORMATION IN TASKS OF INDUCTIVE MODELLING FOR BUSINESS INTELLIGENCE SYSTEMS Artificial Intelligence Methods and Techniques for Business and Engineering Applications 210 INTEGRATED ENVIRONMENT FOR STORING AND HANDLING INFORMATION IN TASKS OF INDUCTIVE MODELLING FOR BUSINESS INTELLIGENCE

More information

Software Quality Characteristics Tested For Mobile Application Development

Software Quality Characteristics Tested For Mobile Application Development Thesis no: MGSE-2015-02 Software Quality Characteristics Tested For Mobile Application Developent Literature Review and Epirical Survey WALEED ANWAR Faculty of Coputing Blekinge Institute of Technology

More information

Work Travel and Decision Probling in the Network Marketing World

Work Travel and Decision Probling in the Network Marketing World TRB Paper No. 03-4348 WORK TRAVEL MODE CHOICE MODELING USING DATA MINING: DECISION TREES AND NEURAL NETWORKS Chi Xie Research Assistant Departent of Civil and Environental Engineering University of Massachusetts,

More information

Research Article Performance Evaluation of Human Resource Outsourcing in Food Processing Enterprises

Research Article Performance Evaluation of Human Resource Outsourcing in Food Processing Enterprises Advance Journal of Food Science and Technology 9(2): 964-969, 205 ISSN: 2042-4868; e-issn: 2042-4876 205 Maxwell Scientific Publication Corp. Subitted: August 0, 205 Accepted: Septeber 3, 205 Published:

More information

A Study on the Chain Restaurants Dynamic Negotiation Games of the Optimization of Joint Procurement of Food Materials

A Study on the Chain Restaurants Dynamic Negotiation Games of the Optimization of Joint Procurement of Food Materials International Journal of Coputer Science & Inforation Technology (IJCSIT) Vol 6, No 1, February 2014 A Study on the Chain estaurants Dynaic Negotiation aes of the Optiization of Joint Procureent of Food

More information

ELECTRIC SERVO MOTOR EQUATIONS AND TIME CONSTANTS

ELECTRIC SERVO MOTOR EQUATIONS AND TIME CONSTANTS ELECIC SEO MOO EQUAIONS AND IME CONSANS George W. Younkin, P.E. Life FELLOW IEEE Industrial Controls Consulting, Div. Bulls Eye Marketing, Inc Fond du c, Wisconsin In the analysis of electric servo drive

More information

Binary Embedding: Fundamental Limits and Fast Algorithm

Binary Embedding: Fundamental Limits and Fast Algorithm Binary Ebedding: Fundaental Liits and Fast Algorith Xinyang Yi The University of Texas at Austin yixy@utexas.edu Eric Price The University of Texas at Austin ecprice@cs.utexas.edu Constantine Caraanis

More information

Markov Models and Their Use for Calculations of Important Traffic Parameters of Contact Center

Markov Models and Their Use for Calculations of Important Traffic Parameters of Contact Center Markov Models and Their Use for Calculations of Iportant Traffic Paraeters of Contact Center ERIK CHROMY, JAN DIEZKA, MATEJ KAVACKY Institute of Telecounications Slovak University of Technology Bratislava

More information

A framework for performance monitoring, load balancing, adaptive timeouts and quality of service in digital libraries

A framework for performance monitoring, load balancing, adaptive timeouts and quality of service in digital libraries Int J Digit Libr (2000) 3: 9 35 INTERNATIONAL JOURNAL ON Digital Libraries Springer-Verlag 2000 A fraework for perforance onitoring, load balancing, adaptive tieouts and quality of service in digital libraries

More information

An improved TF-IDF approach for text classification *

An improved TF-IDF approach for text classification * Zhang et al. / J Zheiang Univ SCI 2005 6A(1:49-55 49 Journal of Zheiang University SCIECE ISS 1009-3095 http://www.zu.edu.cn/zus E-ail: zus@zu.edu.cn An iproved TF-IDF approach for text classification

More information

HOW CLOSE ARE THE OPTION PRICING FORMULAS OF BACHELIER AND BLACK-MERTON-SCHOLES?

HOW CLOSE ARE THE OPTION PRICING FORMULAS OF BACHELIER AND BLACK-MERTON-SCHOLES? HOW CLOSE ARE THE OPTION PRICING FORMULAS OF BACHELIER AND BLACK-MERTON-SCHOLES? WALTER SCHACHERMAYER AND JOSEF TEICHMANN Abstract. We copare the option pricing forulas of Louis Bachelier and Black-Merton-Scholes

More information