HARDWARE SPECIALIZATION OF MACHINE-LEARNING KERNELS: POSSIBILITIES FOR APPLICATIONS AND POSSIBILITIES FOR THE PLATFORM DESIGN SPACE

Size: px
Start display at page:

Download "HARDWARE SPECIALIZATION OF MACHINE-LEARNING KERNELS: POSSIBILITIES FOR APPLICATIONS AND POSSIBILITIES FOR THE PLATFORM DESIGN SPACE"

Transcription

1 HARDWARE SPECIALIZATIO OF MACHIE-LEARIG KERELS: POSSIBILITIES FOR APPLICATIOS AD POSSIBILITIES FOR THE PLATFORM DESIG SPACE (Invted) Kyong Ho Lee, Zhuo Wang, and aveen Verma Prnceton Unversty ABSTRACT Ths paper consders two challengng trends affectng lowpower sensng systems: () the applcatons of nterest ncreasngly nvolve embedded sgnals that are very complex to analyze; and (2) the platforms themselves face elevatng constrants n terms of energy and possbly cost. Motvated by the complextes of analyzng the applcaton sgnals, we emphasze the benefts of data-drven approaches. Most notably, these approaches are based on machne learnng, as opposed to tradtonal DSP. We consder how the algorthms lend themselves to specalzed sgnal-analyss platforms. Hardware specalzaton s wellregarded as an approach to address ssues of computatonal effcency, performance, and capacty, thus playng a key role n leveragng Moore s Law. However, we descrbe how hardware specalzaton of machne-learnng kernels, ths tme wth an explct focus on error reslence, can also play a powerful role n enablng system-wde fault tolerance, thereby adng Moore s Law on another dmenson. Index Terms accelerators, embedded systems hardware reslence, machne learnng.. ITRODUCTIO Generally speakng, the physcal systems that we are nterested n sensng can be extremely complex; these mght nclude physologcal systems n the context of medcal devces [], hgh-value nfrastructure n the context of ndustral montorng [2], or large-scale cvl structures n the context of smart ctes [3]. Constructng analytcal models of the embedded sgnals to enable nferences from sensor data s very often unvable. In such scenaros, datadrven sgnal-analyss technques can play a valuable role [4]. Data-drven technques use sensor data, not only to probe a physcal system of nterest, but also as means of constructng models for sgnal analyss. Such technques have ganed hgh relevance n embedded sensng applcatons for two reasons: () powerful frameworks for data-drven modelng and analyss have recently emerged from the doman of machne learnng, and (2) sensor networks, composed of low-power nodes, have made data, that s potentally hghly nformatve, avalable on a large scale. As a result of these factors, sensng nodes have begun to ncorporate data-drven approaches to solve complex embedded-sgnal-analyss problems [5, 6, etc.]. Asde from analyss of applcaton sgnals, data-drven technques have also shown great promse for addressng modelng and predcton challenges on the platform-desgn and -valdaton levels. Ths tme, the strength of these technques has been n representng a platform s own states, when such states are too complex to model analytcally. Successful applcatons of machne-learnng methods for managng workloads n hgh-performance computng platforms [7], thermal fluctuatons n voltage-performance scalable systems [8], bug exposure n post-slcon valdaton methodologes [9], etc., are all drvng exploratons of how such methods can be more extensvely leveraged. Whle machne-learnng methods appear to have strong relevance to the desgn of low-power embedded systems, n fact they have been nvestgated from the perspectve of ultra-low-power systems on only a very lmted level. The purpose of ths paper s to motvate such a perspectve. Ths nvolves three aspects. The frst aspect s that machnelearnng frameworks have not been formulated wth lowpower system constrants n mnd. As a result, even very smple frameworks can pose lmtng challenges n terms of computatonal energy, memory requrements, etc. The Second aspect s that by precsely consderng how machnelearnng frameworks mght be used wthn sensng applcatons, archtectures for low-power devces can explot hardware specalzaton, yeldng a strong lever for overcomng oppostons to the system constrants. Once sutably ncorporated wthn the hardware archtecture, the thrd aspect s that machne-learnng methods can ad the operaton and desgn-space tradeoffs for the platform tself. In partcular, we look at the urgent challenge of hardware faults, whch can arse due to aggressve manufacturng and/or low-energy crcut-operaton regmes. In both cases, the challenges assocated wth modelng the faults and/or ther manfestatons wthn the embedded data makes machne learnng a potentally powerful approach. 2. MACHIE-LEARIG KERELS FROM THE PERSPECTIVE OF PLATFORM EERGY Machne learnng s a rch doman, offerng a wde range of frameworks. Broadly, the varous frameworks take dfferent approaches to statstcal modelng n an effort to address ether the behavor of the system beng modeled or the manner n whch an algorthm mght use the resultng model. For the purposes of ths paper, we focus on a class of frameworks called dscrmnatve methods. These take a

2 smplstc approach, n that they attempt to model a system as a partcular varable of nterest. Despte ths smplcty, dscrmnatve methods form the bass for powerful classfcaton/regresson algorthms, whch can address a wde range of emergng sensor applcatons [0]. Below, we provde an analyss of the energy lmtatons posed by such frameworks and how these can be addressed ether through mathematcal approxmaton or reformulaton. For analyss and llustraton at the kernel-functon level, we focus on support-vector-machne (SVM) classfers, whch have ganed popularty due to ther computatonal effcency and tranng robustness. The SVM s a dscrmnatve supervsed-learnng framework that performs classfcaton on data, whch has been represented as a feature vector. Durng a tranng phase t uses tranng feature vectors to construct a decson boundary n the feature space that optmally separates regons assocated wth dfferent classes of nterest; though bnary classfcaton s most common, mult-class algorthms can be acheved va varous extensons []. The decson boundary s represented by selectng nstances of tranng feature vectors that le near the edges of the dfferent class-data dstrbutons. These nstances are called the support vectors, and are used to compute the classfcaton decson functon. The challenge that typcally arses s that for the strongest classfcaton kernels, the energy of computng the decson functon scales wth the complexty of the support-vector model. Fg. llustrates ths energy scalng for a radalbass-functon (RBF) kernel, profled on an MSP430 mcroprocessor, wth respect to two dmensons that represent model complexty: () feature-vector dmensonalty, and (2) sze of the support-vector set. As a result, n many applcatons, where hgh-order models are requred, the energy of the kernel domnates. The ponts correspondng to two representatve medcal-sensor applcatons are shown emphaszng the mportance of both dmensons. Fg. : Energy of SVM classfcaton on MSP430 showng scalng wth sze of the support-vector set and the dmensonalty of feature vectors [2]. To address the kernel energy, a varety of approaches have been consdered. In [3] a very low-energy formulaton for smple lnear kernels s used. Eq. gves the SVM decson functon, showng how a lnear kernel permts a factorzaton wheren the computaton over all support vectors s reduced va summaton to a sngle vector (sv s a support vector, x s the nput data s feature vector, and all other parameters are from tranng). Ths overcomes the energy scalng shown n Fg. wth respect to the sze of the support-vector set. Let sv = [ sv sv sv ] and x [ x x x ] = 2 M = 2 K( sv, x)α y b = ( sv x)α y b = = sv α y x b = = w x b The challenge s that n many applcatons, lnear kernels show poor performance compared to non-lnear kernels; non-lnear kernels beneft from the ablty to represent much more flexble decson boundares n the feature space. To enhance the performance of weak, low-energy kernels, a powerful technque s adaptve boostng (adaboost) [4]. It constructs a strong kernel by teratvely combnng weak kernels n a lnear combnaton; the weghts for the lnear combnaton are determned through data-drven tranng. The resultng classfer s substantally more flexble, yeldng low-energy and greatly enhanced performance n many applcatons [5]. Rather than approxmatng a desred decson boundary through boostng, another approach s to approxmate the decson functon assocated wth a strong kernel. In [6], non-lnear SVM kernels, ncludng RBF and sgmod, are mplemented n a logarthmc number system. Ths has the beneft of reducng multplcatons to addtons and subtractons, and, thanks to the reduced precson requrements enabled by logarthmc compresson, the logarthmc transformatons can be realzed va look-up tables and/or pece-wse lnear approxmatons. Although approxmatons can substantally mtgate energy consumpton, they do not overcome the problematc energy scalng wth respect to model complexty shown n Fg.. To address ths an alternate formulaton, that uses ncreased dmensonalty to create a lnear kernel from a non-lnear kernel, has been proposed [2]; as wth the lnear-kernel formulaton n Eq., ths enables factorzaton to overcome energy scalng wth the sze of the support-vector set. Eq. 2 shows the formulaton appled to polynomal kernels. Though these kernels offer only an ntermedate level of flexblty compared to RBF kernels, they have shown to yeld comparable performance for many applcatons, and they substantally outperform lnear kernels [2]. The lnear formulaton s acheved by convertng the support vectors nto matrces. The matrx formulaton has the drawback that energy-scalng wth feature-vector dmensonalty s exacerbated (e.g., scalng s quadratc for a second-order polynomal kernel); however, as descrbed n [7] several generc and applcaton-specfc technques M

3 Let sv = = [ sv sv sv ] and x [ x x x ] 2 M = 2 K( sv, x)α y b 2 = ( β sv x + γ ) α y b = γ sv β = 2 M 2 = sv β M [ x x x ] β sv [ γ β sv β sv β sv ] x α y b γ sv β = 2 M 2 = β sv M [ x x x ] β sv [ γ β sv β sv β sv ] α y x b 2 M 2 M [2nd order polynomal kernel] x 2 M 2 x M x x M Precomputed (M+ x M+ matrx) (2) Fg. 2 shows an accelerator-based mcroprocessor that was proposed for SVM-based sgnal-classfcaton applcatons [8]. The SVM accelerator (SVMA) and actve-learnng data-selecton (ALDS) modules compute classfcaton kernels and tranng-data selecton metrcs for actve learnng [9], respectvely; supportng modules, such as a data-path unt for arthmetc and a CORDIC for non-lnear transformatons, are also ncluded. Arbter exst for dmensonalty reducton. As a result, the formulaton enables substantal energy savngs, between n medcal-sensor applcatons [2]. 3. EXPLOITIG ALGORITHMIC STRUCTURE FOR SPECIALIZATIO The prevous secton noted that a key lmtaton of machnelearnng kernels s how ther energy scales wth model complexty. Ths secton looks at algorthmc structure at the applcaton level to explore archtectures for hardware specalzaton that can address kernel energy. An mportant observaton s that sgnal-analyss algorthms can typcally be dvded nto two parts: () sgnal feature extracton, to sutably represent the data n preparaton for classfcaton; and (2) classfcaton, through the applcaton of a hghorder model. Table shows energy proflng results from two representatve applcatons pertanng to medcal-sgnal analyss. What we see s that datasets requrng hgh-order models for analyss lead to a scenaro wheren classfcaton energy substantally domnates. We note, however, that classfcaton can be acheved through varous kernel functons. Feature extracton, on the other hand, requres a hgh degree of programmablty, as t s closely ted to the applcaton and applcaton sgnals of nterest. Its energy, on the other hand, s far lower. These factors suggest that feature-extracton computatons should preferably be delegated to hghly-programmable hardware (.e., CPUs) whle classfcaton should be delegated to specalzed, energy-effcent co-processors. In an ntegrated mcroprocessor, ths motvates an accelerator-based archtecture. Table. Feature extracton and classfcaton energy; Classfcaton energy domnates over feature extracton. Fg. 2: Archtecture block dagram of an accelerator-based mcroprocessor [8]. An mportant concern wth a hardware-specalzed archtecture, however, s the need for selectve confgurablty. As dscussed n the prevous secton, there exst a wde range of kernels and kernel formatons, leadng to a varety of mplementaton tradeoffs. A noteworthy characterstc of data-drven modelng frameworks s that the optmal desgn-pont parameters can depend strongly on the applcaton data. To enable optmzaton of the desgnpont parameters, the classfcaton accelerator n [8] enables confgurablty of the kernel functon and ts formulaton. Table 2 shows the performance as well as computatonal complexty (represented by cycle count) and memory requrements for two medcal-sensor applcatons over the confguraton space. Fg. 3 shows the de photo and the chp summary along wth the measured energy from two medcal-sensor applcatons. As shown, the use of hardware accelerators dramatcally reduces the overall energy (by 44 and 62, respectvely) thanks to the algorthmc structure of the applcatons. 4. DATA-DRIVE HARDWARE RESILIECE THROUGH MACHIE-LEARIG KERELS An archtecture based on hardware accelerators, as above, s motvated by asymmetry n how specfc computatons mpact the overall energy; namely, hardware specalzaton s benefcal for certan computatons f reducng the energy of those computatons has substantal leverage for reducng the overall energy of the applcaton. In ths secton, we show that wth machne-learnng kernels, a new drver for hardware specalzaton may also be reslence;.e., reslence of machne-learnng kernels can be leveraged to substantally mprove the overall platform reslence.

4 Table 2: Illustraton of performance, computatonalcomplexty, and memory usage of kernels over confgurablty space [8]. Fg. 3: De photo, chp summary, and measured applcaton energy of accelerator-based mcroprocessor [8]. Asymmetry based on reslence characterstcs s n fact a prncple that has been exploted prevously. Prmarly, ths has been n the context of control-flow errors n a processor, whch have a more severe mpact on system reslence compared to data-computaton errors. Recognzng ths has motvated specalzed archtectures and desgn methodologes. In the case of archtectures, ether portons of a mcroprocessor [20] or entre cores [2] have been explctly desgned to be error reslent for handlng control flow, whle the remander of the mcroprocessor or cores have been desgned to have relaxed relablty for data computaton. In the case of desgn methodologes, approaches have been developed wheren the control-crtcal logc paths of a mcroprocessor are syntheszed to have ncreased tmng margn [22]. Asde from control-flow errors, hardware asymmetry has also been used n sgnalprocessng archtectures to effcently explot redundancy through the use of statstcal error correcton. Algorthmc nose tolerance (AT) [23] employs a prmary, fullprecson processor that s allowed to make errors alongsde a reduced-precson estmator. Whle the full-precson processor thus benefts from relaxed desgn constrants, the estmator permts error detecton and correcton, through the the use of the estmator output. Other approaches for explotng redundancy have also been proposed [23]. The avalablty of an error-protected machne-learnng accelerator enables an alternate approach. Below we present an overvew and smulaton results for data-drven hardware reslence (DDHR) [24]. In ths approach, learnng enables adaptatons n a classfcaton decson functon. Ths overcomes the mpact of bt-level errors orgnatng from hardware faults n the feature-extracton processor, but also n all prevous stages (ncludng data-acquston and dataconverson blocks). Thanks to the kernel-functon formulatons and hardware-specalzed archtectures presented prevously, the energy and area costs of a reslent classfer mplementaton can be substantally reduced; DDHR thus targets the hardware assocated wth acqurng and dervng sgnal features, whch are becomng ncreasngly complex and dverse wth the expandng scope of applcatons. 4. Overvew of the Approach Fg. 4 llustrates the concept of DDHR. A key premse of machne-learnng systems s that the data used durng the learnng (tranng) phase must exhbt the same statstcs as the data expected durng the analyss phase. In the approach shown, the tranng features are thus obtaned from the error-prone feature-extracton subsystem. Usng ths data, an error-aware model can be constructed for classfcaton. The error-aware model s meant to model the feature-data varances caused due to the applcaton sgnals, but also those caused due to hardware faults. However, model constructon for a supervsed-learnng classfer requres tranng labels n addton to tranng data. It has been shown [24], that f a temporary error-free detector can be mplemented wthn the archtecture, ths can be used to estmate the tranng labels; though the resultng estmates devate from ground truth, smulaton experments show that they can enable performance very close to that of the errorfree system. It should be noted that the requrement of preservng the sgnal statstcs mples that, generally, only statc faults can be addressed through ths approach; transent fault sources can cause the statstcs to be altered, compromsng the performance of the traned model. Fg. 4: Illustraton of DDHR concept. Temporary systems for modelng and labelng enable constructon of an error-aware model from error-affected tranng data. The error-aware model s then used by a real-tme classfer. Fg. 5 llustrates the mpact of hardware faults on feature data and shows how an error-aware model addresses ths. The data shown s obtaned from gate-level smulatons

5 performed on an EEG-based sezure-detecton processor that has been syntheszed to an ASIC standard-cell lbrary. To smulate faults, the syntheszed netlst has been edted to nsert logc-gate swtchng faults and SRAM bt-cell faults at random. As shown, the orgnal decson boundary, traned usng error-free feature vectors, exhbts poor dscrmnaton of the error-affected sezure and non-sezure feature data. An error-aware model, traned usng the erroraffected feature vectors s also shown, llustratng that good dscrmnaton of the resultng data s now acheved. Fg. 5: EEG-sgnal feature data for a sezure detector (plotted n 2D usng prncple component analyss) showng (a) the error-free dstrbuton and (b) the dstrbuton after ntroducng errors (n 20% of the memory cells). Fg. 6 llustrates the mpact of the error-aware model n restorng the performance of the classfcaton decson functon. The frst hstogram shows the classfer output wth an orgnal model when usng error-free feature data as the nput. As shown, good separaton between sezure and non-sezure classes s observed. The second hstogram shows the mpact of njected faults, llustratng the loss of classfer separablty. Fnally, the thrd hstogram shows the output wth an error-aware model when usng error-affected feature data as the nput. As shown separablty performance close to that of the orgnal system s restored. Fg. 6: Hstograms of the classfer output for a sezure detector (a) wthout errors (baselne), (b) wth memory errors, and (c) wth errors, but usng an error-aware model. 4.2 Analyss and Demonstratons To demonstrate the DDHR concept, two bomedcal sensng applcatons are consdered: electroencephalogram (EEG) based epleptc sezure detecton [25] and electrocardogram (ECG) based arrhythma detecton [26]. The EEG and ECG sgnals used are obtaned from the CHB-MIT Scalp EEG Database and the MIT-BIH Arrhythma Database, respectvely [27]. For both applcatons, the performance metrcs consdered are true postve rate (TP), true negatve rate (T), false postve rate (FP), and false negatve rate (F); these are wdely used metrcs for evaluatng bnary classfcaton. For both applcatons, an RTL descrpton of the featureextracton processor s developed and syntheszed to an ASIC standard-cell lbrary to obtan a gate-level netlst. Two types of hardware faults are then njected at a controlled rate nto the netlst. The frst fault type s SRAM bt-cell faults, wheren the data stored n the SRAM s statcally assgned a value of logc or 0 wth 50% probablty. The second fault type s logc-gate swtchng faults, wheren the output nodes of logc gates are statcally ted to logc or 0 wth 50% probablty. Snce the precse locaton of faults can strongly nfluence the manfestatons of errors, ten nstances of error-njected netlsts are smulated n all cases, wth errors njected at random locatons. Fg. 7 shows the error statstcs of the feature-data derved from gate-level smulatons, llustratng sever mpact of the njected faults. The frst plot shows the magntude of the resultng errors by plottng the RMS value of the errors normalzed to the RMS value of the true features; n many cases, the errors exceed the actual feature values. The second set of plots show the error hstogram dstrbutons for four representatve features; the hstograms exhbt hghly-rregular dstrbutons. Error RMS / Sgnal RMS Dmenson # Fg. 7: (a) RMS error by feature (normalzed to the RMS of the true feature value) and (b) representatve error dstrbutons from four features (cases shown correspond to 20% memory bt-cell errors n the sezure detector). Despte the large error magntudes and rregular error dstrbutons n the resultng feature data, the overall performance of the systems s substantally restored thanks to the use of an error-aware model. Fg. 8 shows the performance (averaged over all ten cases at each fault rate) followng SRAM bt-cell error njecton. As shown, the performance of the orgnal model rapdly degrades, whle that of the error-aware model s essentally restored to errorfree levels. Fg. 8 shows the performance followng logcgate error njecton. The performance exhbts hgher varance for ths fault type; thus results for all ten cases are shown at the respectve fault rates where the performance of the error-aware model begns to degrade. As shown, at these fault rates the orgnal model exhbts degraded performance, whle the error-aware model exhbts restored performance (note, T and F results are not shown as they exhbt no notable degradaton).

6 True Postve (%) SRAM bt-cell Error Rate (%) True Postve (%) SRAM bt-cell Error Rate (%) Fg. 8: Performance followng SRAM error njecton of (a) sezure detector and (b) arrhythma detector. Fg. 9: Performance followng logc-gate error njecton of (a) sezure detector wth fault rate of 0-4 errors/node (0 runs), and (b) arrhythma detector wth fault rate of 7x0-2 per node (0 runs). 5. COCLUSIOS Electroncs for embedded applcatons face a challengng two-sded constrant. On one hand, applcatons are rasng the need to perform analyss over ncreasngly complex sensor data. On the other hand, the hardware platforms themselves are plagued by system-resource lmtatons such as energy consumpton, memory sze, and hardware reslence. Gven the opposng tradeoffs typcally faced by platforms, methods that smultaneously address these constrants can have substantal mpact. Machne-learnng approaches can play such a role by enablng new algorthmc capabltes for handlng applcaton sgnals, whle at the same tme rasng opportuntes for hardware specalzaton due to algorthmc structure. Such specalzaton can be leveraged to not only enable substantal energy savngs, but also to enable system-wde fault tolerance. 6. ACKOWLEDGEMETS Ths work was supported n part by Systems on anoscale Informaton fabrcs (SOIC), one of the sx SRC STARnet Centers, sponsored by MARCO and DARPA. 7. REFERECES. Csavoy, G. Molnar, and T. Denson, Creatng support crcuts for the nervous system: Consderatons for bran-machne nterfacng, VLSI Symp. Crcuts, Jun. 2009, pp G. Sartan and C. Purdy, Transformng a 00-year-old ndustry through analytcs, Teradata. 3. A. Maeda, Technology nnovatons for smart ctes," VLSI Symp. Crcuts, June 202, pp Verma, K. H. Lee, and A. Shoeb, Data-drven approaches for computaton n ntellgent bomedcal devces: A case study of EEG montorng for chronc sezure detecton, J. Low Power Electron. Appl., vol., no., pp , Apr J. Park, et al., A 92mW real-tme traffc sgn recognton system wth robust llumnaton adaptaton and support vector machne, IEEE J. Sold State Crcuts, vol. 47, no., pp , ov Verma, A. Shoeb, J. Bohorquez, J. Dawson, J. Guttag, and A. P. Chandrakasan, A mcropower EEG acquston SoC wth ntegrated feature extracton processor for a chronc sezure detecton system, IEEE J. Sold-State Crcuts, vol. 45, no. 4, pp , Apr A. S. Ganapath, Predctng and optmzng system utlzaton and performance va statstcal machne learnng, Ph.D. dssertaton, Unv. Cal., Berkeley, CA, Y. Ge and Q. Qu, Dynamc thermal management for multmeda applcatons usng machne learnng DAC, Jun. 20, pp A. DeOro, et al., Machne learnng-based anomaly detecton for postslcon bug dagnoss, DATE, March P. Dubey, Recognton, mnng and synthess moves computers to the era of tera, Technology at Intel Magazne, Feb C. W. Hsu and C. J. Ln, A comparson of methods for multclass support vector machnes, IEEE Trans. eural etworks vol.3, no.2, pp , Mar K. H. Lee, S.-Y. Kung, and. Verma, Improvng kernel-energy tradeoffs for machne learnng n mplantable and wearable bomedcal applcatons, ICASSP, May 20, pp D. Anguta, A. Bon, and S. Rdella, A dgtal archtecture for support vector machne: theory, algorthm, and FPGA mplementaton, IEEE Tran. eural etworks, vol 4, no. 5, pp , Sep R. E. Schapre and Y. Freund Boostng: Foundatons and Algorthms, Cambrdge, MA, USA: MIT Press, E. I. Shh, Reducng the computatonal demands of medcal montorng classfers by examnng less data, Ph.D. dssertaton, MIT, Cambrdge, MA, F. M. Khan, M. G. Arnold, and W. M. Pottenger. Fnte precson analyss of support vector machne classfcaton n logarthmc number systems, IEEE EUROMICRO Symp.Dgtal System Desgn (DSD), pp , Sept K. H. Lee, S.-Y. Kung, and. Verma, Low-energy Formulatons of Support Vector Machne Kernel Functons for Bomedcal Sensor Applcatons J. Sgn. Process. Syst., Apr DOI 0.007/s K. H. Lee and. Verma, A V general-purpose bomedcal processor wth confgurable machne-learnng accelerators for hghorder, patent-adaptve montorng, ESSCIRC, 202, pp K. Brnker, Incorporatng dversty n actve learnng wth support vector machnes, n MACHIE LEARIG-ITERATIOAL WORKSHOP THE COFERECE, vol. 20, no., 2003, p Y. Yetm, M. Martonos, and S. Malk, "Extractng useful computaton from error-prone processors for streamng applcatons", DATE, March L. Leem, et al., Error-reslent system archtecture for probablstc applcatons, IEEE/ACM DATE, Mar J. Sartor and R. Kumar, Archtectng Processors to Allow Voltage/Relablty Tradeoffs, CASES, Oct Shanbhag, et al., Stochastc computaton, DAC, June Verma, et al., Enablng system-level platform reslence through embedded data-drven nference capabltes n electronc devces, ICASSP, March A. Shoeb, et al., Applcaton of machne learnng to epleptc sezure detecton, ICML, June E. Ubeyl, ECG beats classfcaton usng multclass support vector machnes wth error correctng output codes, Dgtal Sgnal Processng, May Physoet.

Forecasting the Direction and Strength of Stock Market Movement

Forecasting the Direction and Strength of Stock Market Movement Forecastng the Drecton and Strength of Stock Market Movement Jngwe Chen Mng Chen Nan Ye cjngwe@stanford.edu mchen5@stanford.edu nanye@stanford.edu Abstract - Stock market s one of the most complcated systems

More information

ECE544NA Final Project: Robust Machine Learning Hardware via Classifier Ensemble

ECE544NA Final Project: Robust Machine Learning Hardware via Classifier Ensemble 1 ECE544NA Fnal Project: Robust Machne Learnng Hardware va Classfer Ensemble Sa Zhang, szhang12@llnos.edu Dept. of Electr. & Comput. Eng., Unv. of Illnos at Urbana-Champagn, Urbana, IL, USA Abstract In

More information

What is Candidate Sampling

What is Candidate Sampling What s Canddate Samplng Say we have a multclass or mult label problem where each tranng example ( x, T ) conssts of a context x a small (mult)set of target classes T out of a large unverse L of possble

More information

Feature selection for intrusion detection. Slobodan Petrović NISlab, Gjøvik University College

Feature selection for intrusion detection. Slobodan Petrović NISlab, Gjøvik University College Feature selecton for ntruson detecton Slobodan Petrovć NISlab, Gjøvk Unversty College Contents The feature selecton problem Intruson detecton Traffc features relevant for IDS The CFS measure The mrmr measure

More information

Logistic Regression. Lecture 4: More classifiers and classes. Logistic regression. Adaboost. Optimization. Multiple class classification

Logistic Regression. Lecture 4: More classifiers and classes. Logistic regression. Adaboost. Optimization. Multiple class classification Lecture 4: More classfers and classes C4B Machne Learnng Hlary 20 A. Zsserman Logstc regresson Loss functons revsted Adaboost Loss functons revsted Optmzaton Multple class classfcaton Logstc Regresson

More information

The Development of Web Log Mining Based on Improve-K-Means Clustering Analysis

The Development of Web Log Mining Based on Improve-K-Means Clustering Analysis The Development of Web Log Mnng Based on Improve-K-Means Clusterng Analyss TngZhong Wang * College of Informaton Technology, Luoyang Normal Unversty, Luoyang, 471022, Chna wangtngzhong2@sna.cn Abstract.

More information

Single and multiple stage classifiers implementing logistic discrimination

Single and multiple stage classifiers implementing logistic discrimination Sngle and multple stage classfers mplementng logstc dscrmnaton Hélo Radke Bttencourt 1 Dens Alter de Olvera Moraes 2 Vctor Haertel 2 1 Pontfíca Unversdade Católca do Ro Grande do Sul - PUCRS Av. Ipranga,

More information

A Novel Methodology of Working Capital Management for Large. Public Constructions by Using Fuzzy S-curve Regression

A Novel Methodology of Working Capital Management for Large. Public Constructions by Using Fuzzy S-curve Regression Novel Methodology of Workng Captal Management for Large Publc Constructons by Usng Fuzzy S-curve Regresson Cheng-Wu Chen, Morrs H. L. Wang and Tng-Ya Hseh Department of Cvl Engneerng, Natonal Central Unversty,

More information

A DYNAMIC CRASHING METHOD FOR PROJECT MANAGEMENT USING SIMULATION-BASED OPTIMIZATION. Michael E. Kuhl Radhamés A. Tolentino-Peña

A DYNAMIC CRASHING METHOD FOR PROJECT MANAGEMENT USING SIMULATION-BASED OPTIMIZATION. Michael E. Kuhl Radhamés A. Tolentino-Peña Proceedngs of the 2008 Wnter Smulaton Conference S. J. Mason, R. R. Hll, L. Mönch, O. Rose, T. Jefferson, J. W. Fowler eds. A DYNAMIC CRASHING METHOD FOR PROJECT MANAGEMENT USING SIMULATION-BASED OPTIMIZATION

More information

An Alternative Way to Measure Private Equity Performance

An Alternative Way to Measure Private Equity Performance An Alternatve Way to Measure Prvate Equty Performance Peter Todd Parlux Investment Technology LLC Summary Internal Rate of Return (IRR) s probably the most common way to measure the performance of prvate

More information

Implementation of Deutsch's Algorithm Using Mathcad

Implementation of Deutsch's Algorithm Using Mathcad Implementaton of Deutsch's Algorthm Usng Mathcad Frank Roux The followng s a Mathcad mplementaton of Davd Deutsch's quantum computer prototype as presented on pages - n "Machnes, Logc and Quantum Physcs"

More information

On-Line Fault Detection in Wind Turbine Transmission System using Adaptive Filter and Robust Statistical Features

On-Line Fault Detection in Wind Turbine Transmission System using Adaptive Filter and Robust Statistical Features On-Lne Fault Detecton n Wnd Turbne Transmsson System usng Adaptve Flter and Robust Statstcal Features Ruoyu L Remote Dagnostcs Center SKF USA Inc. 3443 N. Sam Houston Pkwy., Houston TX 77086 Emal: ruoyu.l@skf.com

More information

Fault tolerance in cloud technologies presented as a service

Fault tolerance in cloud technologies presented as a service Internatonal Scentfc Conference Computer Scence 2015 Pavel Dzhunev, PhD student Fault tolerance n cloud technologes presented as a servce INTRODUCTION Improvements n technques for vrtualzaton and performance

More information

Performance Analysis and Coding Strategy of ECOC SVMs

Performance Analysis and Coding Strategy of ECOC SVMs Internatonal Journal of Grd and Dstrbuted Computng Vol.7, No. (04), pp.67-76 http://dx.do.org/0.457/jgdc.04.7..07 Performance Analyss and Codng Strategy of ECOC SVMs Zhgang Yan, and Yuanxuan Yang, School

More information

Module 2 LOSSLESS IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur

Module 2 LOSSLESS IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur Module LOSSLESS IMAGE COMPRESSION SYSTEMS Lesson 3 Lossless Compresson: Huffman Codng Instructonal Objectves At the end of ths lesson, the students should be able to:. Defne and measure source entropy..

More information

INVESTIGATION OF VEHICULAR USERS FAIRNESS IN CDMA-HDR NETWORKS

INVESTIGATION OF VEHICULAR USERS FAIRNESS IN CDMA-HDR NETWORKS 21 22 September 2007, BULGARIA 119 Proceedngs of the Internatonal Conference on Informaton Technologes (InfoTech-2007) 21 st 22 nd September 2007, Bulgara vol. 2 INVESTIGATION OF VEHICULAR USERS FAIRNESS

More information

Open Access A Load Balancing Strategy with Bandwidth Constraint in Cloud Computing. Jing Deng 1,*, Ping Guo 2, Qi Li 3, Haizhu Chen 1

Open Access A Load Balancing Strategy with Bandwidth Constraint in Cloud Computing. Jing Deng 1,*, Ping Guo 2, Qi Li 3, Haizhu Chen 1 Send Orders for Reprnts to reprnts@benthamscence.ae The Open Cybernetcs & Systemcs Journal, 2014, 8, 115-121 115 Open Access A Load Balancng Strategy wth Bandwdth Constrant n Cloud Computng Jng Deng 1,*,

More information

Gender Classification for Real-Time Audience Analysis System

Gender Classification for Real-Time Audience Analysis System Gender Classfcaton for Real-Tme Audence Analyss System Vladmr Khryashchev, Lev Shmaglt, Andrey Shemyakov, Anton Lebedev Yaroslavl State Unversty Yaroslavl, Russa vhr@yandex.ru, shmaglt_lev@yahoo.com, andrey.shemakov@gmal.com,

More information

Politecnico di Torino. Porto Institutional Repository

Politecnico di Torino. Porto Institutional Repository Poltecnco d Torno Porto Insttutonal Repostory [Artcle] A cost-effectve cloud computng framework for acceleratng multmeda communcaton smulatons Orgnal Ctaton: D. Angel, E. Masala (2012). A cost-effectve

More information

Frequency Selective IQ Phase and IQ Amplitude Imbalance Adjustments for OFDM Direct Conversion Transmitters

Frequency Selective IQ Phase and IQ Amplitude Imbalance Adjustments for OFDM Direct Conversion Transmitters Frequency Selectve IQ Phase and IQ Ampltude Imbalance Adjustments for OFDM Drect Converson ransmtters Edmund Coersmeer, Ernst Zelnsk Noka, Meesmannstrasse 103, 44807 Bochum, Germany edmund.coersmeer@noka.com,

More information

How To Solve An Onlne Control Polcy On A Vrtualzed Data Center

How To Solve An Onlne Control Polcy On A Vrtualzed Data Center Dynamc Resource Allocaton and Power Management n Vrtualzed Data Centers Rahul Urgaonkar, Ulas C. Kozat, Ken Igarash, Mchael J. Neely urgaonka@usc.edu, {kozat, garash}@docomolabs-usa.com, mjneely@usc.edu

More information

Damage detection in composite laminates using coin-tap method

Damage detection in composite laminates using coin-tap method Damage detecton n composte lamnates usng con-tap method S.J. Km Korea Aerospace Research Insttute, 45 Eoeun-Dong, Youseong-Gu, 35-333 Daejeon, Republc of Korea yaeln@kar.re.kr 45 The con-tap test has the

More information

Causal, Explanatory Forecasting. Analysis. Regression Analysis. Simple Linear Regression. Which is Independent? Forecasting

Causal, Explanatory Forecasting. Analysis. Regression Analysis. Simple Linear Regression. Which is Independent? Forecasting Causal, Explanatory Forecastng Assumes cause-and-effect relatonshp between system nputs and ts output Forecastng wth Regresson Analyss Rchard S. Barr Inputs System Cause + Effect Relatonshp The job of

More information

An artificial Neural Network approach to monitor and diagnose multi-attribute quality control processes. S. T. A. Niaki*

An artificial Neural Network approach to monitor and diagnose multi-attribute quality control processes. S. T. A. Niaki* Journal of Industral Engneerng Internatonal July 008, Vol. 4, No. 7, 04 Islamc Azad Unversty, South Tehran Branch An artfcal Neural Network approach to montor and dagnose multattrbute qualty control processes

More information

Forecasting the Demand of Emergency Supplies: Based on the CBR Theory and BP Neural Network

Forecasting the Demand of Emergency Supplies: Based on the CBR Theory and BP Neural Network 700 Proceedngs of the 8th Internatonal Conference on Innovaton & Management Forecastng the Demand of Emergency Supples: Based on the CBR Theory and BP Neural Network Fu Deqang, Lu Yun, L Changbng School

More information

L10: Linear discriminants analysis

L10: Linear discriminants analysis L0: Lnear dscrmnants analyss Lnear dscrmnant analyss, two classes Lnear dscrmnant analyss, C classes LDA vs. PCA Lmtatons of LDA Varants of LDA Other dmensonalty reducton methods CSCE 666 Pattern Analyss

More information

A Replication-Based and Fault Tolerant Allocation Algorithm for Cloud Computing

A Replication-Based and Fault Tolerant Allocation Algorithm for Cloud Computing A Replcaton-Based and Fault Tolerant Allocaton Algorthm for Cloud Computng Tork Altameem Dept of Computer Scence, RCC, Kng Saud Unversty, PO Box: 28095 11437 Ryadh-Saud Araba Abstract The very large nfrastructure

More information

Calculation of Sampling Weights

Calculation of Sampling Weights Perre Foy Statstcs Canada 4 Calculaton of Samplng Weghts 4.1 OVERVIEW The basc sample desgn used n TIMSS Populatons 1 and 2 was a two-stage stratfed cluster desgn. 1 The frst stage conssted of a sample

More information

Support Vector Machines

Support Vector Machines Support Vector Machnes Max Wellng Department of Computer Scence Unversty of Toronto 10 Kng s College Road Toronto, M5S 3G5 Canada wellng@cs.toronto.edu Abstract Ths s a note to explan support vector machnes.

More information

Credit Limit Optimization (CLO) for Credit Cards

Credit Limit Optimization (CLO) for Credit Cards Credt Lmt Optmzaton (CLO) for Credt Cards Vay S. Desa CSCC IX, Ednburgh September 8, 2005 Copyrght 2003, SAS Insttute Inc. All rghts reserved. SAS Propretary Agenda Background Tradtonal approaches to credt

More information

Performance Analysis of Energy Consumption of Smartphone Running Mobile Hotspot Application

Performance Analysis of Energy Consumption of Smartphone Running Mobile Hotspot Application Internatonal Journal of mart Grd and lean Energy Performance Analyss of Energy onsumpton of martphone Runnng Moble Hotspot Applcaton Yun on hung a chool of Electronc Engneerng, oongsl Unversty, 511 angdo-dong,

More information

A Crossplatform ECG Compression Library for Mobile HealthCare Services

A Crossplatform ECG Compression Library for Mobile HealthCare Services A Crossplatform ECG Compresson Lbrary for Moble HealthCare Servces Alexander Borodn, Yulya Zavyalova Department of Computer Scence Petrozavodsk State Unversty Petrozavodsk, Russa {aborod, yzavyalo}@cs.petrsu.ru

More information

Lecture 2: Single Layer Perceptrons Kevin Swingler

Lecture 2: Single Layer Perceptrons Kevin Swingler Lecture 2: Sngle Layer Perceptrons Kevn Sngler kms@cs.str.ac.uk Recap: McCulloch-Ptts Neuron Ths vastly smplfed model of real neurons s also knon as a Threshold Logc Unt: W 2 A Y 3 n W n. A set of synapses

More information

RequIn, a tool for fast web traffic inference

RequIn, a tool for fast web traffic inference RequIn, a tool for fast web traffc nference Olver aul, Jean Etenne Kba GET/INT, LOR Department 9 rue Charles Fourer 90 Evry, France Olver.aul@nt-evry.fr, Jean-Etenne.Kba@nt-evry.fr Abstract As networked

More information

Vehicle Detection and Tracking in Video from Moving Airborne Platform

Vehicle Detection and Tracking in Video from Moving Airborne Platform Journal of Computatonal Informaton Systems 10: 12 (2014) 4965 4972 Avalable at http://www.jofcs.com Vehcle Detecton and Trackng n Vdeo from Movng Arborne Platform Lye ZHANG 1,2,, Hua WANG 3, L LI 2 1 School

More information

Statistical Approach for Offline Handwritten Signature Verification

Statistical Approach for Offline Handwritten Signature Verification Journal of Computer Scence 4 (3): 181-185, 2008 ISSN 1549-3636 2008 Scence Publcatons Statstcal Approach for Offlne Handwrtten Sgnature Verfcaton 2 Debnath Bhattacharyya, 1 Samr Kumar Bandyopadhyay, 2

More information

Face Verification Problem. Face Recognition Problem. Application: Access Control. Biometric Authentication. Face Verification (1:1 matching)

Face Verification Problem. Face Recognition Problem. Application: Access Control. Biometric Authentication. Face Verification (1:1 matching) Face Recognton Problem Face Verfcaton Problem Face Verfcaton (1:1 matchng) Querymage face query Face Recognton (1:N matchng) database Applcaton: Access Control www.vsage.com www.vsoncs.com Bometrc Authentcaton

More information

The OC Curve of Attribute Acceptance Plans

The OC Curve of Attribute Acceptance Plans The OC Curve of Attrbute Acceptance Plans The Operatng Characterstc (OC) curve descrbes the probablty of acceptng a lot as a functon of the lot s qualty. Fgure 1 shows a typcal OC Curve. 10 8 6 4 1 3 4

More information

An Interest-Oriented Network Evolution Mechanism for Online Communities

An Interest-Oriented Network Evolution Mechanism for Online Communities An Interest-Orented Network Evoluton Mechansm for Onlne Communtes Cahong Sun and Xaopng Yang School of Informaton, Renmn Unversty of Chna, Bejng 100872, P.R. Chna {chsun,yang}@ruc.edu.cn Abstract. Onlne

More information

DEFINING %COMPLETE IN MICROSOFT PROJECT

DEFINING %COMPLETE IN MICROSOFT PROJECT CelersSystems DEFINING %COMPLETE IN MICROSOFT PROJECT PREPARED BY James E Aksel, PMP, PMI-SP, MVP For Addtonal Informaton about Earned Value Management Systems and reportng, please contact: CelersSystems,

More information

On the Optimal Control of a Cascade of Hydro-Electric Power Stations

On the Optimal Control of a Cascade of Hydro-Electric Power Stations On the Optmal Control of a Cascade of Hydro-Electrc Power Statons M.C.M. Guedes a, A.F. Rbero a, G.V. Smrnov b and S. Vlela c a Department of Mathematcs, School of Scences, Unversty of Porto, Portugal;

More information

An ILP Formulation for Task Mapping and Scheduling on Multi-core Architectures

An ILP Formulation for Task Mapping and Scheduling on Multi-core Architectures An ILP Formulaton for Task Mappng and Schedulng on Mult-core Archtectures Yng Y, We Han, Xn Zhao, Ahmet T. Erdogan and Tughrul Arslan Unversty of Ednburgh, The Kng's Buldngs, Mayfeld Road, Ednburgh, EH9

More information

Institute of Informatics, Faculty of Business and Management, Brno University of Technology,Czech Republic

Institute of Informatics, Faculty of Business and Management, Brno University of Technology,Czech Republic Lagrange Multplers as Quanttatve Indcators n Economcs Ivan Mezník Insttute of Informatcs, Faculty of Busness and Management, Brno Unversty of TechnologCzech Republc Abstract The quanttatve role of Lagrange

More information

J. Parallel Distrib. Comput.

J. Parallel Distrib. Comput. J. Parallel Dstrb. Comput. 71 (2011) 62 76 Contents lsts avalable at ScenceDrect J. Parallel Dstrb. Comput. journal homepage: www.elsever.com/locate/jpdc Optmzng server placement n dstrbuted systems n

More information

An Analysis of Central Processor Scheduling in Multiprogrammed Computer Systems

An Analysis of Central Processor Scheduling in Multiprogrammed Computer Systems STAN-CS-73-355 I SU-SE-73-013 An Analyss of Central Processor Schedulng n Multprogrammed Computer Systems (Dgest Edton) by Thomas G. Prce October 1972 Techncal Report No. 57 Reproducton n whole or n part

More information

SIMULATION OPTIMIZATION: APPLICATIONS IN RISK MANAGEMENT

SIMULATION OPTIMIZATION: APPLICATIONS IN RISK MANAGEMENT Internatonal Journal of Informaton Technology & Decson Makng Vol. 7, No. 4 (2008) 571 587 c World Scentfc Publshng Company SIMULATION OPTIMIZATION: APPLICATIONS IN RISK MANAGEMENT MARCO BETTER and FRED

More information

Improved SVM in Cloud Computing Information Mining

Improved SVM in Cloud Computing Information Mining Internatonal Journal of Grd Dstrbuton Computng Vol.8, No.1 (015), pp.33-40 http://dx.do.org/10.1457/jgdc.015.8.1.04 Improved n Cloud Computng Informaton Mnng Lvshuhong (ZhengDe polytechnc college JangSu

More information

BERNSTEIN POLYNOMIALS

BERNSTEIN POLYNOMIALS On-Lne Geometrc Modelng Notes BERNSTEIN POLYNOMIALS Kenneth I. Joy Vsualzaton and Graphcs Research Group Department of Computer Scence Unversty of Calforna, Davs Overvew Polynomals are ncredbly useful

More information

Modelling of Web Domain Visits by Radial Basis Function Neural Networks and Support Vector Machine Regression

Modelling of Web Domain Visits by Radial Basis Function Neural Networks and Support Vector Machine Regression Modellng of Web Doman Vsts by Radal Bass Functon Neural Networks and Support Vector Machne Regresson Vladmír Olej, Jana Flpová Insttute of System Engneerng and Informatcs Faculty of Economcs and Admnstraton,

More information

RESEARCH ON DUAL-SHAKER SINE VIBRATION CONTROL. Yaoqi FENG 1, Hanping QIU 1. China Academy of Space Technology (CAST) yaoqi.feng@yahoo.

RESEARCH ON DUAL-SHAKER SINE VIBRATION CONTROL. Yaoqi FENG 1, Hanping QIU 1. China Academy of Space Technology (CAST) yaoqi.feng@yahoo. ICSV4 Carns Australa 9- July, 007 RESEARCH ON DUAL-SHAKER SINE VIBRATION CONTROL Yaoq FENG, Hanpng QIU Dynamc Test Laboratory, BISEE Chna Academy of Space Technology (CAST) yaoq.feng@yahoo.com Abstract

More information

Can Auto Liability Insurance Purchases Signal Risk Attitude?

Can Auto Liability Insurance Purchases Signal Risk Attitude? Internatonal Journal of Busness and Economcs, 2011, Vol. 10, No. 2, 159-164 Can Auto Lablty Insurance Purchases Sgnal Rsk Atttude? Chu-Shu L Department of Internatonal Busness, Asa Unversty, Tawan Sheng-Chang

More information

Calculating the high frequency transmission line parameters of power cables

Calculating the high frequency transmission line parameters of power cables < ' Calculatng the hgh frequency transmsson lne parameters of power cables Authors: Dr. John Dcknson, Laboratory Servces Manager, N 0 RW E B Communcatons Mr. Peter J. Ncholson, Project Assgnment Manager,

More information

Pre-allocation Strategies of Computational Resources in Cloud Computing using Adaptive Resonance Theory-2

Pre-allocation Strategies of Computational Resources in Cloud Computing using Adaptive Resonance Theory-2 Pre-allocaton Strateges of Computatonal Resources n Cloud Computng usng Adaptve Resonance Theory-2 Dr.T. R. Gopalakrshnan Nar 1, P Jayarekha 2 1 Drector, Research and Industry Incubaton Centre(RIIC), DSI,

More information

A Suspect Vehicle Tracking System Based on Video

A Suspect Vehicle Tracking System Based on Video 3rd Internatonal Conference on Multmeda Technology ICMT 2013) A Suspect Vehcle Trackng System Based on Vdeo Yad Chen 1, Tuo Wang Abstract. Vdeo survellance systems are wdely used n securty feld. The large

More information

Traffic State Estimation in the Traffic Management Center of Berlin

Traffic State Estimation in the Traffic Management Center of Berlin Traffc State Estmaton n the Traffc Management Center of Berln Authors: Peter Vortsch, PTV AG, Stumpfstrasse, D-763 Karlsruhe, Germany phone ++49/72/965/35, emal peter.vortsch@ptv.de Peter Möhl, PTV AG,

More information

ECONOMICS OF PLANT ENERGY SAVINGS PROJECTS IN A CHANGING MARKET Douglas C White Emerson Process Management

ECONOMICS OF PLANT ENERGY SAVINGS PROJECTS IN A CHANGING MARKET Douglas C White Emerson Process Management ECONOMICS OF PLANT ENERGY SAVINGS PROJECTS IN A CHANGING MARKET Douglas C Whte Emerson Process Management Abstract Energy prces have exhbted sgnfcant volatlty n recent years. For example, natural gas prces

More information

An Evaluation of the Extended Logistic, Simple Logistic, and Gompertz Models for Forecasting Short Lifecycle Products and Services

An Evaluation of the Extended Logistic, Simple Logistic, and Gompertz Models for Forecasting Short Lifecycle Products and Services An Evaluaton of the Extended Logstc, Smple Logstc, and Gompertz Models for Forecastng Short Lfecycle Products and Servces Charles V. Trappey a,1, Hsn-yng Wu b a Professor (Management Scence), Natonal Chao

More information

THE DISTRIBUTION OF LOAN PORTFOLIO VALUE * Oldrich Alfons Vasicek

THE DISTRIBUTION OF LOAN PORTFOLIO VALUE * Oldrich Alfons Vasicek HE DISRIBUION OF LOAN PORFOLIO VALUE * Oldrch Alfons Vascek he amount of captal necessary to support a portfolo of debt securtes depends on the probablty dstrbuton of the portfolo loss. Consder a portfolo

More information

Detecting Credit Card Fraud using Periodic Features

Detecting Credit Card Fraud using Periodic Features Detectng Credt Card Fraud usng Perodc Features Alejandro Correa Bahnsen, Djamla Aouada, Aleksandar Stojanovc and Björn Ottersten Interdscplnary Centre for Securty, Relablty and Trust Unversty of Luxembourg,

More information

Methodology to Determine Relationships between Performance Factors in Hadoop Cloud Computing Applications

Methodology to Determine Relationships between Performance Factors in Hadoop Cloud Computing Applications Methodology to Determne Relatonshps between Performance Factors n Hadoop Cloud Computng Applcatons Lus Eduardo Bautsta Vllalpando 1,2, Alan Aprl 1 and Alan Abran 1 1 Department of Software Engneerng and

More information

An MILP model for planning of batch plants operating in a campaign-mode

An MILP model for planning of batch plants operating in a campaign-mode An MILP model for plannng of batch plants operatng n a campagn-mode Yanna Fumero Insttuto de Desarrollo y Dseño CONICET UTN yfumero@santafe-concet.gov.ar Gabrela Corsano Insttuto de Desarrollo y Dseño

More information

Efficient QoS Aggregation in Service Value Networks

Efficient QoS Aggregation in Service Value Networks 22 45th Hawa Internatonal Conference on System Scences Effcent QoS Aggregaton n Servce Value etworks Steffen Haak Research Center for Informaton Technology (FZI) haak@fz.de Benjamn Blau SAP AG benjamn.blau@sap.com

More information

行 政 院 國 家 科 學 委 員 會 補 助 專 題 研 究 計 畫 成 果 報 告 期 中 進 度 報 告

行 政 院 國 家 科 學 委 員 會 補 助 專 題 研 究 計 畫 成 果 報 告 期 中 進 度 報 告 行 政 院 國 家 科 學 委 員 會 補 助 專 題 研 究 計 畫 成 果 報 告 期 中 進 度 報 告 畫 類 別 : 個 別 型 計 畫 半 導 體 產 業 大 型 廠 房 之 設 施 規 劃 計 畫 編 號 :NSC 96-2628-E-009-026-MY3 執 行 期 間 : 2007 年 8 月 1 日 至 2010 年 7 月 31 日 計 畫 主 持 人 : 巫 木 誠 共 同

More information

J. Parallel Distrib. Comput. Environment-conscious scheduling of HPC applications on distributed Cloud-oriented data centers

J. Parallel Distrib. Comput. Environment-conscious scheduling of HPC applications on distributed Cloud-oriented data centers J. Parallel Dstrb. Comput. 71 (2011) 732 749 Contents lsts avalable at ScenceDrect J. Parallel Dstrb. Comput. ournal homepage: www.elsever.com/locate/pdc Envronment-conscous schedulng of HPC applcatons

More information

Portfolio Loss Distribution

Portfolio Loss Distribution Portfolo Loss Dstrbuton Rsky assets n loan ortfolo hghly llqud assets hold-to-maturty n the bank s balance sheet Outstandngs The orton of the bank asset that has already been extended to borrowers. Commtment

More information

Application of Multi-Agents for Fault Detection and Reconfiguration of Power Distribution Systems

Application of Multi-Agents for Fault Detection and Reconfiguration of Power Distribution Systems 1 Applcaton of Mult-Agents for Fault Detecton and Reconfguraton of Power Dstrbuton Systems K. Nareshkumar, Member, IEEE, M. A. Choudhry, Senor Member, IEEE, J. La, A. Felach, Senor Member, IEEE Abstract--The

More information

Course outline. Financial Time Series Analysis. Overview. Data analysis. Predictive signal. Trading strategy

Course outline. Financial Time Series Analysis. Overview. Data analysis. Predictive signal. Trading strategy Fnancal Tme Seres Analyss Patrck McSharry patrck@mcsharry.net www.mcsharry.net Trnty Term 2014 Mathematcal Insttute Unversty of Oxford Course outlne 1. Data analyss, probablty, correlatons, vsualsaton

More information

Enabling P2P One-view Multi-party Video Conferencing

Enabling P2P One-view Multi-party Video Conferencing Enablng P2P One-vew Mult-party Vdeo Conferencng Yongxang Zhao, Yong Lu, Changja Chen, and JanYn Zhang Abstract Mult-Party Vdeo Conferencng (MPVC) facltates realtme group nteracton between users. Whle P2P

More information

benefit is 2, paid if the policyholder dies within the year, and probability of death within the year is ).

benefit is 2, paid if the policyholder dies within the year, and probability of death within the year is ). REVIEW OF RISK MANAGEMENT CONCEPTS LOSS DISTRIBUTIONS AND INSURANCE Loss and nsurance: When someone s subject to the rsk of ncurrng a fnancal loss, the loss s generally modeled usng a random varable or

More information

When Network Effect Meets Congestion Effect: Leveraging Social Services for Wireless Services

When Network Effect Meets Congestion Effect: Leveraging Social Services for Wireless Services When Network Effect Meets Congeston Effect: Leveragng Socal Servces for Wreless Servces aowen Gong School of Electrcal, Computer and Energy Engeerng Arzona State Unversty Tempe, AZ 8587, USA xgong9@asuedu

More information

IMPACT ANALYSIS OF A CELLULAR PHONE

IMPACT ANALYSIS OF A CELLULAR PHONE 4 th ASA & μeta Internatonal Conference IMPACT AALYSIS OF A CELLULAR PHOE We Lu, 2 Hongy L Bejng FEAonlne Engneerng Co.,Ltd. Bejng, Chna ABSTRACT Drop test smulaton plays an mportant role n nvestgatng

More information

METHODOLOGY TO DETERMINE RELATIONSHIPS BETWEEN PERFORMANCE FACTORS IN HADOOP CLOUD COMPUTING APPLICATIONS

METHODOLOGY TO DETERMINE RELATIONSHIPS BETWEEN PERFORMANCE FACTORS IN HADOOP CLOUD COMPUTING APPLICATIONS METHODOLOGY TO DETERMINE RELATIONSHIPS BETWEEN PERFORMANCE FACTORS IN HADOOP CLOUD COMPUTING APPLICATIONS Lus Eduardo Bautsta Vllalpando 1,2, Alan Aprl 1 and Alan Abran 1 1 Department of Software Engneerng

More information

Efficient Striping Techniques for Variable Bit Rate Continuous Media File Servers æ

Efficient Striping Techniques for Variable Bit Rate Continuous Media File Servers æ Effcent Strpng Technques for Varable Bt Rate Contnuous Meda Fle Servers æ Prashant J. Shenoy Harrck M. Vn Department of Computer Scence, Department of Computer Scences, Unversty of Massachusetts at Amherst

More information

Analysis of Energy-Conserving Access Protocols for Wireless Identification Networks

Analysis of Energy-Conserving Access Protocols for Wireless Identification Networks From the Proceedngs of Internatonal Conference on Telecommuncaton Systems (ITC-97), March 2-23, 1997. 1 Analyss of Energy-Conservng Access Protocols for Wreless Identfcaton etworks Imrch Chlamtac a, Chara

More information

Vision Mouse. Saurabh Sarkar a* University of Cincinnati, Cincinnati, USA ABSTRACT 1. INTRODUCTION

Vision Mouse. Saurabh Sarkar a* University of Cincinnati, Cincinnati, USA ABSTRACT 1. INTRODUCTION Vson Mouse Saurabh Sarkar a* a Unversty of Cncnnat, Cncnnat, USA ABSTRACT The report dscusses a vson based approach towards trackng of eyes and fngers. The report descrbes the process of locatng the possble

More information

A hybrid global optimization algorithm based on parallel chaos optimization and outlook algorithm

A hybrid global optimization algorithm based on parallel chaos optimization and outlook algorithm Avalable onlne www.ocpr.com Journal of Chemcal and Pharmaceutcal Research, 2014, 6(7):1884-1889 Research Artcle ISSN : 0975-7384 CODEN(USA) : JCPRC5 A hybrd global optmzaton algorthm based on parallel

More information

Survey on Virtual Machine Placement Techniques in Cloud Computing Environment

Survey on Virtual Machine Placement Techniques in Cloud Computing Environment Survey on Vrtual Machne Placement Technques n Cloud Computng Envronment Rajeev Kumar Gupta and R. K. Paterya Department of Computer Scence & Engneerng, MANIT, Bhopal, Inda ABSTRACT In tradtonal data center

More information

Comparison of Control Strategies for Shunt Active Power Filter under Different Load Conditions

Comparison of Control Strategies for Shunt Active Power Filter under Different Load Conditions Comparson of Control Strateges for Shunt Actve Power Flter under Dfferent Load Condtons Sanjay C. Patel 1, Tushar A. Patel 2 Lecturer, Electrcal Department, Government Polytechnc, alsad, Gujarat, Inda

More information

Cloud-based Social Application Deployment using Local Processing and Global Distribution

Cloud-based Social Application Deployment using Local Processing and Global Distribution Cloud-based Socal Applcaton Deployment usng Local Processng and Global Dstrbuton Zh Wang *, Baochun L, Lfeng Sun *, and Shqang Yang * * Bejng Key Laboratory of Networked Multmeda Department of Computer

More information

Network Aware Load-Balancing via Parallel VM Migration for Data Centers

Network Aware Load-Balancing via Parallel VM Migration for Data Centers Network Aware Load-Balancng va Parallel VM Mgraton for Data Centers Kun-Tng Chen 2, Chen Chen 12, Po-Hsang Wang 2 1 Informaton Technology Servce Center, 2 Department of Computer Scence Natonal Chao Tung

More information

Quantization Effects in Digital Filters

Quantization Effects in Digital Filters Quantzaton Effects n Dgtal Flters Dstrbuton of Truncaton Errors In two's complement representaton an exact number would have nfntely many bts (n general). When we lmt the number of bts to some fnte value

More information

Adaptive Sampling for Energy Conservation in Wireless Sensor Networks for Snow Monitoring Applications

Adaptive Sampling for Energy Conservation in Wireless Sensor Networks for Snow Monitoring Applications Adaptve Samplng for Energy Conservaton n Wreless Sensor Networks for Snow Montorng Applcatons Cesare Alpp *, Guseppe Anastas, Crstan Galpert *, Francesca Mancn, Manuel Rover * * Dp. d Elettronca e Informazone

More information

Eye Center Localization on a Facial Image Based on Multi-Block Local Binary Patterns

Eye Center Localization on a Facial Image Based on Multi-Block Local Binary Patterns Eye Center Localzaton on a Facal Image Based on Mult-Bloc Local Bnary Patterns Anatoly tn, Vladmr Khryashchev, Olga Stepanova Yaroslavl State Unversty Yaroslavl, Russa anatolyntnyar@gmal.com, vhr@yandex.ru,

More information

A Hierarchical Anomaly Network Intrusion Detection System using Neural Network Classification

A Hierarchical Anomaly Network Intrusion Detection System using Neural Network Classification IDC IDC A Herarchcal Anomaly Network Intruson Detecton System usng Neural Network Classfcaton ZHENG ZHANG, JUN LI, C. N. MANIKOPOULOS, JAY JORGENSON and JOSE UCLES ECE Department, New Jersey Inst. of Tech.,

More information

Project Networks With Mixed-Time Constraints

Project Networks With Mixed-Time Constraints Project Networs Wth Mxed-Tme Constrants L Caccetta and B Wattananon Western Australan Centre of Excellence n Industral Optmsaton (WACEIO) Curtn Unversty of Technology GPO Box U1987 Perth Western Australa

More information

Optimal resource capacity management for stochastic networks

Optimal resource capacity management for stochastic networks Submtted for publcaton. Optmal resource capacty management for stochastc networks A.B. Deker H. Mlton Stewart School of ISyE, Georga Insttute of Technology, Atlanta, GA 30332, ton.deker@sye.gatech.edu

More information

MACHINE VISION SYSTEM FOR SPECULAR SURFACE INSPECTION: USE OF SIMULATION PROCESS AS A TOOL FOR DESIGN AND OPTIMIZATION

MACHINE VISION SYSTEM FOR SPECULAR SURFACE INSPECTION: USE OF SIMULATION PROCESS AS A TOOL FOR DESIGN AND OPTIMIZATION MACHINE VISION SYSTEM FOR SPECULAR SURFACE INSPECTION: USE OF SIMULATION PROCESS AS A TOOL FOR DESIGN AND OPTIMIZATION R. SEULIN, F. MERIENNE and P. GORRIA Laboratore Le2, CNRS FRE2309, EA 242, Unversté

More information

ESTABLISHING TRADE-OFFS BETWEEN SUSTAINED AND MOMENTARY RELIABILITY INDICES IN ELECTRIC DISTRIBUTION PROTECTION DESIGN: A GOAL PROGRAMMING APPROACH

ESTABLISHING TRADE-OFFS BETWEEN SUSTAINED AND MOMENTARY RELIABILITY INDICES IN ELECTRIC DISTRIBUTION PROTECTION DESIGN: A GOAL PROGRAMMING APPROACH ESTABLISHIG TRADE-OFFS BETWEE SUSTAIED AD MOMETARY RELIABILITY IDICES I ELECTRIC DISTRIBUTIO PROTECTIO DESIG: A GOAL PROGRAMMIG APPROACH Gustavo D. Ferrera, Arturo S. Bretas, Maro O. Olvera Federal Unversty

More information

Development of an intelligent system for tool wear monitoring applying neural networks

Development of an intelligent system for tool wear monitoring applying neural networks of Achevements n Materals and Manufacturng Engneerng VOLUME 14 ISSUE 1-2 January-February 2006 Development of an ntellgent system for tool wear montorng applyng neural networks A. Antć a, J. Hodolč a,

More information

Sketching Sampled Data Streams

Sketching Sampled Data Streams Sketchng Sampled Data Streams Florn Rusu, Aln Dobra CISE Department Unversty of Florda Ganesvlle, FL, USA frusu@cse.ufl.edu adobra@cse.ufl.edu Abstract Samplng s used as a unversal method to reduce the

More information

THE APPLICATION OF DATA MINING TECHNIQUES AND MULTIPLE CLASSIFIERS TO MARKETING DECISION

THE APPLICATION OF DATA MINING TECHNIQUES AND MULTIPLE CLASSIFIERS TO MARKETING DECISION Internatonal Journal of Electronc Busness Management, Vol. 3, No. 4, pp. 30-30 (2005) 30 THE APPLICATION OF DATA MINING TECHNIQUES AND MULTIPLE CLASSIFIERS TO MARKETING DECISION Yu-Mn Chang *, Yu-Cheh

More information

A Design Method of High-availability and Low-optical-loss Optical Aggregation Network Architecture

A Design Method of High-availability and Low-optical-loss Optical Aggregation Network Architecture A Desgn Method of Hgh-avalablty and Low-optcal-loss Optcal Aggregaton Network Archtecture Takehro Sato, Kuntaka Ashzawa, Kazumasa Tokuhash, Dasuke Ish, Satoru Okamoto and Naoak Yamanaka Dept. of Informaton

More information

Towards Specialization of the Contract-Aware Software Development Process

Towards Specialization of the Contract-Aware Software Development Process Towards Specalzaton of the Contract-Aware Software Development Process Anna Derezńska, Przemysław Ołtarzewsk Insttute of Computer Scence, Warsaw Unversty of Technology, Nowowejska 5/9, 00-665 Warsaw, Poland

More information

Time Domain simulation of PD Propagation in XLPE Cables Considering Frequency Dependent Parameters

Time Domain simulation of PD Propagation in XLPE Cables Considering Frequency Dependent Parameters Internatonal Journal of Smart Grd and Clean Energy Tme Doman smulaton of PD Propagaton n XLPE Cables Consderng Frequency Dependent Parameters We Zhang a, Jan He b, Ln Tan b, Xuejun Lv b, Hong-Je L a *

More information

ANALYZING THE RELATIONSHIPS BETWEEN QUALITY, TIME, AND COST IN PROJECT MANAGEMENT DECISION MAKING

ANALYZING THE RELATIONSHIPS BETWEEN QUALITY, TIME, AND COST IN PROJECT MANAGEMENT DECISION MAKING ANALYZING THE RELATIONSHIPS BETWEEN QUALITY, TIME, AND COST IN PROJECT MANAGEMENT DECISION MAKING Matthew J. Lberatore, Department of Management and Operatons, Vllanova Unversty, Vllanova, PA 19085, 610-519-4390,

More information

Network Security Situation Evaluation Method for Distributed Denial of Service

Network Security Situation Evaluation Method for Distributed Denial of Service Network Securty Stuaton Evaluaton Method for Dstrbuted Denal of Servce Jn Q,2, Cu YMn,2, Huang MnHuan,2, Kuang XaoHu,2, TangHong,2 ) Scence and Technology on Informaton System Securty Laboratory, Bejng,

More information

ONE of the most crucial problems that every image

ONE of the most crucial problems that every image IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 23, NO. 10, OCTOBER 2014 4413 Maxmum Margn Projecton Subspace Learnng for Vsual Data Analyss Symeon Nktds, Anastasos Tefas, Member, IEEE, and Ioanns Ptas, Fellow,

More information

Cost Minimization using Renewable Cooling and Thermal Energy Storage in CDNs

Cost Minimization using Renewable Cooling and Thermal Energy Storage in CDNs Cost Mnmzaton usng Renewable Coolng and Thermal Energy Storage n CDNs Stephen Lee College of Informaton and Computer Scences UMass, Amherst stephenlee@cs.umass.edu Rahul Urgaonkar IBM Research rurgaon@us.bm.com

More information

An Efficient and Simplified Model for Forecasting using SRM

An Efficient and Simplified Model for Forecasting using SRM HAFIZ MUHAMMAD SHAHZAD ASIF*, MUHAMMAD FAISAL HAYAT*, AND TAUQIR AHMAD* RECEIVED ON 15.04.013 ACCEPTED ON 09.01.014 ABSTRACT Learnng form contnuous fnancal systems play a vtal role n enterprse operatons.

More information

Enterprise Master Patient Index

Enterprise Master Patient Index Enterprse Master Patent Index Healthcare data are captured n many dfferent settngs such as hosptals, clncs, labs, and physcan offces. Accordng to a report by the CDC, patents n the Unted States made an

More information