Learning from Multiple Related Data Streams with Asynchronous Flowing Speeds

Size: px
Start display at page:

Download "Learning from Multiple Related Data Streams with Asynchronous Flowing Speeds"

Transcription

1 Learnng from Multple Related Data Streams wth Asynchronous Flowng Speeds Zh Qao, Peng Zhang, Jng He, Jnghua Yan, L Guo Insttute of Computng Technology, Chnese Academy of Scences, Bejng, , Chna. School of Engneerng & Scence, Vctora Unversty, PO Box 14428, Melbourne, Australa qaozh@software.ct.ac.cn, Jng.He@vu.edu.au, {zhangpeng, Abstract Related data streams refer to data streams that can be joned together by matchng ther jon attrbutes. Exstng research on learnng from related data streams s based on an assumpton that all streams arrve at a central processng unt n a synchronous way, such that n an arbtrary sldng wndow, all tuples of the streams can be perfectly joned together. Ths assumpton, however, does not hold when related data streams are generated or transferred at dfferent speeds, and thus may arrve n the central processng unt n an asynchronous manner. In ths paper, we argue that for asynchronous data streams, there exst a small porton of perfectly joned examples (.e., complete examples) and a large porton of partally joned examples (.e., ncomplete examples). Accordngly, we present a new Learnng from Complete and Fxed Examples () framework that can fx ncomplete examples to boost the learnng. Experments on both synthetc and real-world data streams demonstrate that s able to acheve a hgher predcton accuracy for learnng from related data streams than other smple solutons can offer. I. INTRODUCTION Exstng work n data stream mnng has made great efforts n knowledge dscovery for a sngle data stream [1, 2, 8, 9], but fndng patterns from multple related data streams s stll nadequately addressed. For many real-world data stream applcatons, stream data are often collected from dfferent channels wth dfferent modaltes. Under such envronments, t s natural to combne multple data streams together to dscover trs and patterns underneath the stream data. Related data streams refer to data streams that can be joned together by some shared jon attrbutes. In ths paper, we consder the problem of learnng from multple related data streams wth asynchronous speeds. Learnng from multple related data streams has been dscussed before, but manly from the prvacy preservng data stream mnng [3] perspectve. In ther studes, related data streams are assumed flowng synchronously, such that stream data can be perfectly joned together n a sldng wndow through ther jon attrbutes. However, n many real-world data stream applcatons, related data streams may be generated or transferred at dfferent speeds, and thus may arrve at the central processng unt n an asynchronous way. As a result, the jon attrbutes may not match each other perfectly n a sldng wndow. Under ths observaton, our man goal n ths paper s to learn from multple related data streams by takng the asynchronous factor nto consderaton, whch, to the best of our knowledge, has not been addressed before. To make the concept clear, assume the gven of two examples n Fgures 1 and 2, both of whch are a snapshot of a sldng wndow (Fgure 1 s a wdely used example n the prvacy preservng data stream mnng area [3]). In the examples, the target s to dscover patterns of the proftable tradng n stock markets by combnng related data streams, such as the phone call streams between dealers and managers/staffs of publc companes, the stock tradng streams that record the tradng actons, and the News streams from a local TV staton reportng ther forecasts on the stock markets. From Fgure 1, t s clear that all the jon attrbutes (whch are denoted by the arrows connectng the streams) are perfectly matched n the snapshot. Ths, however, only happens n an deal stuaton. A much more common case, as shown n Fgure 2, s that jon attrbutes may not be perfectly matched n a sldng wndow. For nstance, at the tme pont 9:02am, Denns gves Peter a call tellng hm to sell CCC s stock, but n the tradng stream we can t get Ross s matchng tuples (whch may not be avalable untl 9:04am due to the communcaton delay). As a result, the jon operaton wll generate few perfectly joned examples as shown n Table 1 (whch are referred to as the complete examples) and a large porton of partally joned examples as shown n Table 2 (whch are referred to as the ncomplete examples n ths paper). Learnng from multple related data streams should, therefore, take both types of examples nto consderaton. In ths paper, we present a Learnng from Complete and Fxed Examples () method for learnng from multple related data stream wth asynchronous speeds. The am of the model s to fx the large porton of ncomplete examples usng nformaton from complete examples to boost the learnng. Experments on both synthetc and real world data streams demonstrate that s able to help buld models wth a hgher predcton accuracy than other smple solutons can offer. The rest of ths paper s structured as follows. In the next secton, we formulate the problem of learnng from multple related data streams and dscuss some smple solutons. We ntroduce the learnng framework n Secton 3, and conduct experments on both synthetc and real world data streams n Secton 4. We survey related work n Secton 5, and conclude the paper n Secton 6.

2 Table I THE PERFECTLY JOINED EXAMPLES (WHICH ARE REFERRED TO AS complete examples) Org Caller Callee Dealer Type Stock Company Class Stock Forecast AAA Adams Jack Jack Sell A AAA Yes A Yes BBB Ray Selna Selna Buy B BBB Yes B No Table II THE PARTIALLY JOINED EXAMPLES (WHICH ARE REFERRED TO AS ncomplete examples) Org Caller Callee Dealer Type Stock Company Class Stock Forecast CCC Denns Peter?????????? John Buy C CCC No C No??? Ross Sell E EEE Yes?????????? D Yes Org Caller Callee 9:00 AAA Adams Jack 9:01 BBB Ray Selna 9:02 CCC Denns Peter Phone Call Stream Dealer Type Stock Comp. Class 9:01 Jack Sell A AAA Yes 9:02 Selna Buy B BBB Yes 9:03 Peter Sell A AAA No 9:03 Peter Buy C CCC Yes Tradng Stream Stock Fore. 9:00 A Yes 9:01 B No 9:02 C No Ne ws Stre am Fgure 1. An llustraton of multple related data streams wth synchronous speeds. Org Caller Callee 9:00 AAA Adams Jack 9:01 BBB Ray Selna 9:02 CCC Denns Peter Phone Call Stream Dealer Type Stock Comp. Class 9:01 Jack Sell A AAA Yes 9:02 Selna Buy B BBB Yes 9:02 John Buy C CCC No 9:03 Ross Sell E EEE Yes Tra dng Stream Stock Fore. 9:00 A Yes 9:01 B No 9:02 C No 9:03 D Yes Ne ws Stre am Fgure 2. An llustraton of multple related data streams wth asynchronous speeds. II. PROBLEM DEFINITION AND SIMPLE SOLUTIONS Consder m related data streams {S 1, S 2,...,S m }. All the streams share some jon attrbutes and can be joned together accordng to certan condtons. The joned stream has n classes {c 1, c 2,...,c n }. The sldng wndow sze s set to be w. A complete example s generated by jonng m tuples from m streams, whereas an ncomplete example s generated by jonng r (1 r < m) tuples from r streams. For an ncomplete example, we dvde t nto three parts: a set of observed attrbutes O, a set of unobserved attrbutes U, and ts class label L. Moreover, observed attrbutes O can be further dvded nto {o 1, o 2,...,o r }, where o s a tuple from stream S. Besdes, n ths paper we assume that all the attrbutes are ndepent wth each other. A. Problem Defnton Learnng from multple related data streams s not a trval task, consderng that even learnng from one sngle data stream s severely challenged by data volumes and concept drftng realtes. In multple related data streams scenaro, dfferent streams may be generated at dfferent stes that are far away from each other, whch makes related tuples may arrve at the central processng unt at dfferent speeds. Therefore, after the jon operaton, the generated tranng examples are lkely to contan a small porton of complete examples and a large porton of ncomplete examples. Such realtes n asynchronous data streams envronment rase the followng concerns. Frst, snce the complete examples provde the global vew of the related data streams, makng proper use of such type of examples s a basc necessty. Second, snce the ncomplete examples reflect local nformaton of the data dstrbuton, usng them properly to boost the performance plays an mportant role n buldng an accurate model. Thrd, tuples whch can not be perfectly joned n the current sldng wndow does not mean that they are useless, consderng that ther related tuples n other streams may arrve n succeedng wndows. Fourth, patterns and trs from multple related data streams may experence concept drftng wth tme elapsng. To sum up, learnng from multple related data streams needs to take the followng four concerns nto consderaton: Be able to make proper use of the complete examples whch provde the global vew of the multple related data streams; Be able to make use of the ncomplete examples properly to boost the performance. In asynchronous data streams, ncomplete examples take a large porton of the tranng examples. Any mproper use of such data may worsen the performance. Be able to ncrementally update the hstorcal nformaton over data streams. Be able to handle concept drftng. When a new concept emerges, adaptng quckly to ft the new concept s a necessty. B. Smple Solutons Intutvely, the followng two methods can be appled to learn from multple related data streams. Learn from Complete Examples (): drops all the ncomplete examples, and only learns from the complete

3 examples. For nstance, as far as Fgure 2 s consdered, just uses complete examples n Table 1 to buld model, but dscards all the ncomplete examples n Table 2. The mert of s that t usually has a low executon overhead because t only uses a small porton of the complete examples to buld model. However, ts lmtaton s also apparent. Snce the complete examples are sparse, t usually can not yeld a good predcton model. Learn from Complete and Incomplete Examples () Unlke the method that smply drops all the ncomplete examples, keeps all the ncomplete examples. To the unobserved attrbutes n the ncomplete examples, marks them wth a unform symbol? as shown n Table 2. By dong so, the orgnal problem s converted to a new problem of learnng wth unobserved values. The mert of s that t not only uses complete examples, but also uses ncomplete examples to buld model. When the complete examples are sparse, s able to acheve satsfactory results by consderng the ncomplete examples. The lmtaton of s that t flls up all the unobserved values smply wth a unform symbol?, whch can not always acheve a satsfactory result. A better alternatve method s to use the hstorcal nformaton from the passed stream data to fx the unobserved values, as what we wll do n the followng Leanng from Complete and Fxed Examples () method. III. FORMULATION OF THE MODEL In ths secton, we descrbe the formulaton of the model n detal. As we dscussed above, t s a common case that multple related data streams arrve at the central processng unt n an asynchronous manner, such that only a small porton of tuples n a sldng wndow can be completely joned, leavng a large porton of tuples be ncompletely joned. Thus, the generated tranng examples usually contan only few complete examples and a lot of ncomplete examples. The goal of our model s to summarze all the hstorcal complete records to nfer the mssng values n the ncomplete examples to boost the performance. A. Learn from Complete Examples Complete examples provde a global vew of related data streams, whch s valuable for understandng the patterns and trs behnd the streams. In ths secton, we consder how to summarze the completed examples over data streams. In data streams, t s mpractcal to buffer all the hstorcal data to boost the performance. In order to get hstorcal nformaton wthout bufferng all the data, an alternatve way s to desgn lght weght data structures to summarze nformaton from the passed data. Ths knd of data structures should be able to ncrementally update when new data comes, meanwhle, t also should be able to handle concept drftng problem. A well known example of such data structures s the mcro-cluster structure proposed for clusterng data streams n [4]. In ths paper, we use the sample average as the basc data structure to summarze the data streams. More precsely, for class label c, we defne a sample average vector X c = 1 S c S c S c j as the basc data structure, where S c denotes the total number of complete examples n the passed data havng class label c, and S c j denotes the j th example n the passed data havng class label c. On one hand, X c s able to ncrementally mantan the hstorcal nformaton over data streams. For nstance, when a new sldng wndow comes, the sample average X c can be updated as n Eq.(1), X c = X c X c = S c S c + S c X S c + c S c + S c X c (1) where X c s the sample average of all the complete examples wth class label c n the new sldng wndow, and S c denotes the total number of complete examples havng class label c. From Eq. (1), we can observe that X c s able to scale up over large amounts of data. On the other hand, the sample average X c also can be used to detect and handle concept drftng problem. Gven a parameter λ ( λ>0), concept drftng can be defned as that the current sample average X c bases from the mantaned sample average X c larger than λ. Snce concept drftng wll make the kept X c be too obsolete to reflect the current data dstrbuton, we wll reset X c usng the current sample average X c (.e., let X c = X c ) when the concept drfts. By dong so, t s safe to say that mantanng the sample average of the complete examples s an effectve way to handle both ncremental learnng and concept drftng problems. B. Learn from Incomplete Examples Although ncomplete examples provde less nformaton than the completed examples, they should not be neglected for two reasons. Frst, ncomplete examples usually take up a large porton of the whole tranng examples, and any mproper use of such data may even deterorate the performance. Second, ncomplete examples can provde useful local nformaton to some extent. Therefore, learnng from such type of examples properly s also very mportant. Unlke the method whch smply marks all the unobserved values wth?, our method tres to fx the unobserved values usng X c. For an ncomplete example, unobserved values exst n two dfferent types: (1) only some attrbutes are unobserved (.e., the 2 nd and 3 th examples n Table 2), and (2) both some attrbutes and the class label are unobserved (.e., the 1 th and 4 th examples n Table 2). To a Type (1) example, we frst use ts observed attrbutes to update the hstorcal nformaton X c, and then use the correspondng subvector n X c to fll up ts unobserved attrbutes. To a Type (2) example, snce t has both unobserved attrbutes and unobserved class label, we gve a two-stage learnng strategy to fx t as descrbed n Theorem 1.

4 Theorem 1 Assume all the attrbutes are ndepent. If we want to fx an ncomplete example havng both unobserved attrbutes and unobserved class label, we should frst fx ts class label accordng to the observed attrbutes, and then fx ts unobserved attrbutes accordng to the fxed class label. Proof The goal s to fx both the unobserved label L and the unobserved attrbutes U usng the observed attrbutes O. Accordng to the probablty theory, P(U, L, O) P(L, O)P(U L, O) P(U, L O) = = P(O) P(O) = P(O)P(L O)P(U L) = P(L O)P(U L) (2) P(O) Thus, n order to fx the unobserved L and U, we should frst fx L usng P(L O), and then fx U usng P(U L). After formulatng ths two-stage learnng method, a followng queston s how to calculate the probabltes P(L O) and P(U L). As far as P(L O) s consdered, accordng to the Bayesan decson rule, the class label L should be the one whch has the maxmal probablty P(c O) (1 n), P(L O) = max P(c O). (3) Snce all the attrbutes are ndepent, Eq.(3) can be further transformed to Eq.(4), P(L O) = max P(c O) = max P(c o j ). (4) =1 Eq.(4) shows that each observed tuple o j ( j = 1,...,r) wll put a weght to decde the fnal class label L. Therefore, to fx an ncomplete example s label, t s essental to take all ts observed attrbutes nto consderaton. By usng the logarthm functon, we transform Eq.(4) to Eq.(5) as follow, P(L O) max r log P(c o j ) max P(c o j ). (5) Eq. (5) shows that the fnal class label L wll be the class c that has the largest probablty r log P(c o j ). Snce the probabltes P(c o j )( = 1,...,c; j = 1,...,r) are hard to calculate, especally when there exst contnuous attrbute values, we use an approxmate method named Label Consensus Score to approxmately calculate every P(c o j ). Defnton 1(Label Consensus Score) The label consensus score between label c and an observed tuple o j s defned as the dstance value wth reverse proporton to the Eucldean dstance between the sample average X j and tuple o j. The fnal class label L wll be the class c whch has the largest label consensus score on all the r observed tuples as shown n Eq. (6), P(L O) max P(c o j ) max score(c, o j ) (6) After gettng the class label L, P(U L) can be easly calculated by usng the hstorcal nformaton X L to fll up all the unobserved attrbutes U. C. The Learnng Framework Algorthm 1 shows the learnng framework, whch conssts of fve major steps. In the frst step, jons all the tuples n the current sldng wndow to get a tranng examples T. Note that T contans a small porton of completed examples and a large porton of ncomplete examples. In the second step, the completed examples are used to update the hstorcal nformaton over data streams. More precsely, t frst calculates the sample average X c on each class label c n the current sldng wndow, and then compares X c wth the hstorcal one X c. If the dfference between them s larger than a gven threshold λ, then t replaces X c wth the current one X c; otherwse, t absorbs X c nto the X c.in the thrd step, the ncomplete examples are fxed by fllng up all the unobserved values. If the class label c s observed, then t uses all the observed attrbutes o j to update the X c j ; otherwse, the unobserved class label s estmated usng Eq.(6). Besdes, all the unobserved attrbutes are fxed usng the hstorcal nformaton from X c. In the fourth step, the completed examples and the fxed examples are combned together to buld a classfcaton model. In the last step, the model s tested on a next sldng wndow. IV. EXPERIMENTS In ths secton, we report expermental results and comparsons of the proposed framework from the followng three aspects: performance wth respect to (1) dfferent sldng wndow szes (.e., w), (2) dfferent success jon rates (.e., p), whch s used to smulate the dfferent arrvng speeds among dfferent data streams, and (3) dfferent concept drftng thresholds (.e., λ ). Benchmark methods: To assess s performance, we use the and methods as the benchmark methods for comparsons. As dscussed above, learns only from the completed examples, whle learns from both the completed and ncomplete examples. All these learnng methods are mplemented n Java wth an ntegraton of WEKA data mnng tool [5], and the Lbsvm (The java package can be downloaded from yasser/wlsvm/.) s used as the base classfcaton model. Synthetc Data Streams: We employ the followng four steps to generate synthetc related data streams wth asynchronous speeds. Frstly, we generate a sngle large data stream S. Then we vertcally splt S nto several equal ntervals. After that, each nterval s assgned an addtonal jon attrbute td, and treated as a sngle data stream. Fnally, we use p to smulate the asynchronous streams scenaro (.e., for an arbtrary tuple t, we generate a random number rd (0 < rd < 1). If rd < p, then t wll be taken as an unobserved tuple). More specfcally, we frst desgn a sequence of pars S = {(x 1, y 1 ),...,(x T, y T )}, where x R 19 s the attrbutes and y { 1, +1} s the class label. The classfcaton boundary s defned as 19 =1 a x = a 0, and the concept

5 Input : Related data streams S 1,...,S m, sldng wndow w, concept drftng λ Output: Average accuracy P and varance V Randomly generate a X c for each class label c /* Intalzaton */; for each sldng wndow w do T=jon streams(s 1,...,S m ) /* Step 1 */; for the completed examples do /* Step 2 */ for each class label c do f X c X c λ then X c X c X c ; else X c = X c for all ncomplete examples do /* Step 3 */ f class label c s observed then foreach observed tuple o j ( j = 1,...,r) do X c j X c j o j ; else Calculate the class label usng Eq. (6) ; Fx all the unobserved attrbutes usng X c ; Buld up a classfer f on the completed & fxed examples /* Step 4 */; Test classfer f on the next sldng wndow, and get P /* Step 5 */; Output the average accuracy P and the varance V. Algorthm 1: The learnng framework drftng s smulated by makng every a ( = 0,...,19) have 10% chance to evolve to a and 5% chance to reverse ts drecton. After that, we splt S vertcally and equally nto fve ntervals, wth each nterval havng fve attrbutes. At last, we use dfferent p to smulate the asynchronous streams scenaro. Real World Data Streams: We use the URL Reputaton data streams from the UCI Machne Learnng Repostory [6]. The goal s to detect malcous web stes by combnng the host-based features and the lexcal features of ther URLs. Fgure 3 llustrates the data collecton archtecture. The malcous URLs are obtaned from a large Web mal provder, whle the bengn URLs are randomly drawn from Yahoo s drecton lstng. For every ncomng URL, the feature collector collects the URL s host-based features by queryng DNS, WHOIS, blacklst and geographc nformaton servers, and collects the lexcal features from the lexcal related servers. Snce these servers locate at dfferent stes, we can take these related data streams as asynchronous data streams. We also use the parameter p here to smulate the Fgure 3. Overvew of real-tme URL feed, feature collecton, and classfcaton nfrastructure [7]. dfferent speeds among each server. We wll analyss the frst week data n our experments. For each type of attrbutes, we extract the former 20 attrbutes for analyss. Expermental Results: We lst our expermental results n Fgure 4 and Table 3. The parameters, f not specally mentoned, are set as follows: λ = 0.3, w = 500, and p = 0.5. Fgure 4(a) shows comparson results wth respect to dfferent sldng wndow szes. It s obvous that always performs the best, especally when the wndow sze s small. Ths s because under asynchronous data streams scenaro, the smaller the wndow sze s, the less chance to get completed tranng examples. Fortunately, s able to receve good predcton accuraces by fxng the ncomplete examples among asynchronous data streams. To further explore the relatonshp between the asynchronous rate p and the predcton accuracy, we conduct another experment as shown n Fgure 4(b). Not surprsngly, all the three algorthms suffer a loss when the number of ncomplete examples ncreases. But s able to avod a sgnfcant drop by fllng up the unobserved attrbutes usng the hstorcal sample average nformaton. In Fgure 4(c), we test dfferent values of the parameter λ whch defnes the concept drftng threshold. As shown n the fgure, performs the best when λ = 0.3, and thus n followng experments, we let λ be 0.3. We lst the expermental results on the real world URL Reputaton data streams n Table 3. All the three models are compared on the frst week data streams wth respect to ther predcton accuraces and computaton overheads. The sldng wndow sze s set to be 500, so there are totally 40 sldng wndows everyday. From Table 3, we can observe that always has the best predcton accuraces, follows as the second best model, and s the least accurate model. Ths valdates our argument that by fxng the ncomplete examples usng the hstorcal nformaton, s able to acheve good results n asynchronous data streams. Fgure 4(e) shows the detaled comparsons on the 40 sldng wndows on a specfc day (.e., Day 4). It s clear that always has the best predcton accuraces over all the 40 wndows. Consequently, t s safe to say that s able to learn accurately and tmely from related data streams wth asynchronous arrvng speeds.

6 Wndow Sze (a) Parameter w Parameter P (b) Parameter p Parameter λ (c) Parameter λ Tme Cost (Mllsecond) Wndow Sze Wndow ID (d) Tme cost (e) Chunk-by-chunk comparsons on Day 4 Fg. 4. Comparson results on both synthetc and real world data streams Table III COMPARISONS OF ACCURACY AND TIME COST (MILLISECOND) ON THE URL REPUTATION DATA. Day1 298 ± ± ±0.093 Day2 010 ± ± ±0.066 Day3 320 ± ± ±0.085 Day4 100 ± ± ±0.074 Day5 410 ± ± ±0.093 Day6 303 ± ± ±0.092 Day ± ± ±0.050 V. CONCLUSIONS In ths paper, we consder a new problem of learnng from multple related asynchronous data streams. We frst argued that to learn from such data streams, four concerns should be taken nto consderaton: (1) use the complete examples to gan a global vew across multple data streams, (2) make proper use of the ncomplete examples to boost the learnng, (3) ncrementally mantan the hstorcal nformaton from the hstorcal data, and (4) detect and handle concept drftng n data streams. To meet these challenges, n ths paper we presented a new Learnng from Complete and Fxed Examples () framework to learn from multple related data streams wth dfferent flowng speeds. More specfcally, frst combnes all the related data streams to generate tranng examples n the current sldng wndow, whch may contan a small porton of complete examples and a large porton of ncomplete examples. After that, employs a two stage method to fx the ncomplete examples by fxng the class label and the unobserved attrbutes. At last, bulds a predcton model on the completed examples. Durng the whole learnng procedure, the sample average s used to ncrementally mantan the hstorcal nformaton. The concept drftng problem s also addressed by measurng the change of the sample average. The contrbuton of work reported n the paper s fourfold: (1) we frst consder the problem of learnng from multple related data streams wth asynchronous speeds, and formulate ths problem as learnng from complete and ncomplete examples; (2) we propose a two stage learnng method to fx the unobserved class label and attrbutes for ncomplete examples; (3) we propose an effcent label consensus score method to approxmately calculate the Bayesan decson rule on data streams; (4) we propose a learnng framework to learn from multple related data streams wth asynchronous speeds. ACKNOWLEDGMENT Ths research was partally supported by the Natonal Scence Foundaton of Chna (NSFC) under Grant No , and Basc Research Program of Chna (973 Program) under Grant No.2007CB REFERENCES [1] K. Crammer, M. Kearns, J. Wortman: Learnng from Multpel Sources. Journal of Machne Learnng Research 9 (2008). [2] P. Zhang, X. Zhu, L. Guo: Mnng Data Streams wth Labeled and Unlabeled Tranng Examples. In: Proc. of IEEE ICDM 09 (2009) [3] Y. Xu, K. Wang, A. Fu, R. She, J. Pe: Classfcaton Spannng Correlated Data Streams. In Proc. of CIKM 2006, (2005). [4] C. Aggarwal et al.: A framework for clusterng evolvng data streams. In: Proceedngs of VLDB (2003). [5] I. Wtten, and E. Frank: Data mnng: practcal machne learnng tools and technques, Morgan Kaufmann (2005). [6] D. Newman, S. Hettch, C. Blake, and C Merz: UCI Repostory of machne learnng (1998). [7] J. Ma et al.: Identfyng Suspcous URLs: An Applcaton of Large-Scale Onlne Learnng, In Proc. of ICML 09 (2009). [8] P. Zhang, X. Zhu, Y. Sh: Categorzng and Mnng Concept Drftng Data Streams, In: Proc. of KDD 08 (2008). [9] X. Zhu, P. Zhang, X. Ln, Y. Sh: Actve Learnng from Stream Data Usng Optmal Weght Classfer Ensemble. IEEE Trans. on System, Man, Cybernetcs, Part B, Vol. 40 (4) (2010)

Learning the Kernel Parameters in Kernel Minimum Distance Classifier

Learning the Kernel Parameters in Kernel Minimum Distance Classifier Learnng the Kernel Parameters n Kernel Mnmum Dstance Classfer Daoqang Zhang 1,, Songcan Chen and Zh-Hua Zhou 1* 1 Natonal Laboratory for Novel Software Technology Nanjng Unversty, Nanjng 193, Chna Department

More information

A Binarization Algorithm specialized on Document Images and Photos

A Binarization Algorithm specialized on Document Images and Photos A Bnarzaton Algorthm specalzed on Document mages and Photos Ergna Kavalleratou Dept. of nformaton and Communcaton Systems Engneerng Unversty of the Aegean kavalleratou@aegean.gr Abstract n ths paper, a

More information

Parallelism for Nested Loops with Non-uniform and Flow Dependences

Parallelism for Nested Loops with Non-uniform and Flow Dependences Parallelsm for Nested Loops wth Non-unform and Flow Dependences Sam-Jn Jeong Dept. of Informaton & Communcaton Engneerng, Cheonan Unversty, 5, Anseo-dong, Cheonan, Chungnam, 330-80, Korea. seong@cheonan.ac.kr

More information

Machine Learning: Algorithms and Applications

Machine Learning: Algorithms and Applications 14/05/1 Machne Learnng: Algorthms and Applcatons Florano Zn Free Unversty of Bozen-Bolzano Faculty of Computer Scence Academc Year 011-01 Lecture 10: 14 May 01 Unsupervsed Learnng cont Sldes courtesy of

More information

Cluster Analysis of Electrical Behavior

Cluster Analysis of Electrical Behavior Journal of Computer and Communcatons, 205, 3, 88-93 Publshed Onlne May 205 n ScRes. http://www.scrp.org/ournal/cc http://dx.do.org/0.4236/cc.205.350 Cluster Analyss of Electrcal Behavor Ln Lu Ln Lu, School

More information

Helsinki University Of Technology, Systems Analysis Laboratory Mat Independent research projects in applied mathematics (3 cr)

Helsinki University Of Technology, Systems Analysis Laboratory Mat Independent research projects in applied mathematics (3 cr) Helsnk Unversty Of Technology, Systems Analyss Laboratory Mat-2.08 Independent research projects n appled mathematcs (3 cr) "! #$&% Antt Laukkanen 506 R ajlaukka@cc.hut.f 2 Introducton...3 2 Multattrbute

More information

BOOSTING CLASSIFICATION ACCURACY WITH SAMPLES CHOSEN FROM A VALIDATION SET

BOOSTING CLASSIFICATION ACCURACY WITH SAMPLES CHOSEN FROM A VALIDATION SET 1 BOOSTING CLASSIFICATION ACCURACY WITH SAMPLES CHOSEN FROM A VALIDATION SET TZU-CHENG CHUANG School of Electrcal and Computer Engneerng, Purdue Unversty, West Lafayette, Indana 47907 SAUL B. GELFAND School

More information

BAYESIAN MULTI-SOURCE DOMAIN ADAPTATION

BAYESIAN MULTI-SOURCE DOMAIN ADAPTATION BAYESIAN MULTI-SOURCE DOMAIN ADAPTATION SHI-LIANG SUN, HONG-LEI SHI Department of Computer Scence and Technology, East Chna Normal Unversty 500 Dongchuan Road, Shangha 200241, P. R. Chna E-MAIL: slsun@cs.ecnu.edu.cn,

More information

BioTechnology. An Indian Journal FULL PAPER. Trade Science Inc.

BioTechnology. An Indian Journal FULL PAPER. Trade Science Inc. [Type text] [Type text] [Type text] ISSN : 0974-74 Volume 0 Issue BoTechnology 04 An Indan Journal FULL PAPER BTAIJ 0() 04 [684-689] Revew on Chna s sports ndustry fnancng market based on market -orented

More information

Classifier Selection Based on Data Complexity Measures *

Classifier Selection Based on Data Complexity Measures * Classfer Selecton Based on Data Complexty Measures * Edth Hernández-Reyes, J.A. Carrasco-Ochoa, and J.Fco. Martínez-Trndad Natonal Insttute for Astrophyscs, Optcs and Electroncs, Lus Enrque Erro No.1 Sta.

More information

Content Based Image Retrieval Using 2-D Discrete Wavelet with Texture Feature with Different Classifiers

Content Based Image Retrieval Using 2-D Discrete Wavelet with Texture Feature with Different Classifiers IOSR Journal of Electroncs and Communcaton Engneerng (IOSR-JECE) e-issn: 78-834,p- ISSN: 78-8735.Volume 9, Issue, Ver. IV (Mar - Apr. 04), PP 0-07 Content Based Image Retreval Usng -D Dscrete Wavelet wth

More information

6.854 Advanced Algorithms Petar Maymounkov Problem Set 11 (November 23, 2005) With: Benjamin Rossman, Oren Weimann, and Pouya Kheradpour

6.854 Advanced Algorithms Petar Maymounkov Problem Set 11 (November 23, 2005) With: Benjamin Rossman, Oren Weimann, and Pouya Kheradpour 6.854 Advanced Algorthms Petar Maymounkov Problem Set 11 (November 23, 2005) Wth: Benjamn Rossman, Oren Wemann, and Pouya Kheradpour Problem 1. We reduce vertex cover to MAX-SAT wth weghts, such that the

More information

Mathematics 256 a course in differential equations for engineering students

Mathematics 256 a course in differential equations for engineering students Mathematcs 56 a course n dfferental equatons for engneerng students Chapter 5. More effcent methods of numercal soluton Euler s method s qute neffcent. Because the error s essentally proportonal to the

More information

User Authentication Based On Behavioral Mouse Dynamics Biometrics

User Authentication Based On Behavioral Mouse Dynamics Biometrics User Authentcaton Based On Behavoral Mouse Dynamcs Bometrcs Chee-Hyung Yoon Danel Donghyun Km Department of Computer Scence Department of Computer Scence Stanford Unversty Stanford Unversty Stanford, CA

More information

Term Weighting Classification System Using the Chi-square Statistic for the Classification Subtask at NTCIR-6 Patent Retrieval Task

Term Weighting Classification System Using the Chi-square Statistic for the Classification Subtask at NTCIR-6 Patent Retrieval Task Proceedngs of NTCIR-6 Workshop Meetng, May 15-18, 2007, Tokyo, Japan Term Weghtng Classfcaton System Usng the Ch-square Statstc for the Classfcaton Subtask at NTCIR-6 Patent Retreval Task Kotaro Hashmoto

More information

Steps for Computing the Dissimilarity, Entropy, Herfindahl-Hirschman and. Accessibility (Gravity with Competition) Indices

Steps for Computing the Dissimilarity, Entropy, Herfindahl-Hirschman and. Accessibility (Gravity with Competition) Indices Steps for Computng the Dssmlarty, Entropy, Herfndahl-Hrschman and Accessblty (Gravty wth Competton) Indces I. Dssmlarty Index Measurement: The followng formula can be used to measure the evenness between

More information

Problem Set 3 Solutions

Problem Set 3 Solutions Introducton to Algorthms October 4, 2002 Massachusetts Insttute of Technology 6046J/18410J Professors Erk Demane and Shaf Goldwasser Handout 14 Problem Set 3 Solutons (Exercses were not to be turned n,

More information

TECHNIQUE OF FORMATION HOMOGENEOUS SAMPLE SAME OBJECTS. Muradaliyev A.Z.

TECHNIQUE OF FORMATION HOMOGENEOUS SAMPLE SAME OBJECTS. Muradaliyev A.Z. TECHNIQUE OF FORMATION HOMOGENEOUS SAMPLE SAME OBJECTS Muradalyev AZ Azerbajan Scentfc-Research and Desgn-Prospectng Insttute of Energetc AZ1012, Ave HZardab-94 E-mal:aydn_murad@yahoocom Importance of

More information

Lecture 5: Multilayer Perceptrons

Lecture 5: Multilayer Perceptrons Lecture 5: Multlayer Perceptrons Roger Grosse 1 Introducton So far, we ve only talked about lnear models: lnear regresson and lnear bnary classfers. We noted that there are functons that can t be represented

More information

An Optimal Algorithm for Prufer Codes *

An Optimal Algorithm for Prufer Codes * J. Software Engneerng & Applcatons, 2009, 2: 111-115 do:10.4236/jsea.2009.22016 Publshed Onlne July 2009 (www.scrp.org/journal/jsea) An Optmal Algorthm for Prufer Codes * Xaodong Wang 1, 2, Le Wang 3,

More information

Simulation Based Analysis of FAST TCP using OMNET++

Simulation Based Analysis of FAST TCP using OMNET++ Smulaton Based Analyss of FAST TCP usng OMNET++ Umar ul Hassan 04030038@lums.edu.pk Md Term Report CS678 Topcs n Internet Research Sprng, 2006 Introducton Internet traffc s doublng roughly every 3 months

More information

A Fast Visual Tracking Algorithm Based on Circle Pixels Matching

A Fast Visual Tracking Algorithm Based on Circle Pixels Matching A Fast Vsual Trackng Algorthm Based on Crcle Pxels Matchng Zhqang Hou hou_zhq@sohu.com Chongzhao Han czhan@mal.xjtu.edu.cn Ln Zheng Abstract: A fast vsual trackng algorthm based on crcle pxels matchng

More information

Related-Mode Attacks on CTR Encryption Mode

Related-Mode Attacks on CTR Encryption Mode Internatonal Journal of Network Securty, Vol.4, No.3, PP.282 287, May 2007 282 Related-Mode Attacks on CTR Encrypton Mode Dayn Wang, Dongda Ln, and Wenlng Wu (Correspondng author: Dayn Wang) Key Laboratory

More information

Fast Computation of Shortest Path for Visiting Segments in the Plane

Fast Computation of Shortest Path for Visiting Segments in the Plane Send Orders for Reprnts to reprnts@benthamscence.ae 4 The Open Cybernetcs & Systemcs Journal, 04, 8, 4-9 Open Access Fast Computaton of Shortest Path for Vstng Segments n the Plane Ljuan Wang,, Bo Jang

More information

The Research of Support Vector Machine in Agricultural Data Classification

The Research of Support Vector Machine in Agricultural Data Classification The Research of Support Vector Machne n Agrcultural Data Classfcaton Le Sh, Qguo Duan, Xnmng Ma, Me Weng College of Informaton and Management Scence, HeNan Agrcultural Unversty, Zhengzhou 45000 Chna Zhengzhou

More information

Compiler Design. Spring Register Allocation. Sample Exercises and Solutions. Prof. Pedro C. Diniz

Compiler Design. Spring Register Allocation. Sample Exercises and Solutions. Prof. Pedro C. Diniz Compler Desgn Sprng 2014 Regster Allocaton Sample Exercses and Solutons Prof. Pedro C. Dnz USC / Informaton Scences Insttute 4676 Admralty Way, Sute 1001 Marna del Rey, Calforna 90292 pedro@s.edu Regster

More information

Outline. Type of Machine Learning. Examples of Application. Unsupervised Learning

Outline. Type of Machine Learning. Examples of Application. Unsupervised Learning Outlne Artfcal Intellgence and ts applcatons Lecture 8 Unsupervsed Learnng Professor Danel Yeung danyeung@eee.org Dr. Patrck Chan patrckchan@eee.org South Chna Unversty of Technology, Chna Introducton

More information

Scheduling Remote Access to Scientific Instruments in Cyberinfrastructure for Education and Research

Scheduling Remote Access to Scientific Instruments in Cyberinfrastructure for Education and Research Schedulng Remote Access to Scentfc Instruments n Cybernfrastructure for Educaton and Research Je Yn 1, Junwe Cao 2,3,*, Yuexuan Wang 4, Lanchen Lu 1,3 and Cheng Wu 1,3 1 Natonal CIMS Engneerng and Research

More information

Module Management Tool in Software Development Organizations

Module Management Tool in Software Development Organizations Journal of Computer Scence (5): 8-, 7 ISSN 59-66 7 Scence Publcatons Management Tool n Software Development Organzatons Ahmad A. Al-Rababah and Mohammad A. Al-Rababah Faculty of IT, Al-Ahlyyah Amman Unversty,

More information

Determining the Optimal Bandwidth Based on Multi-criterion Fusion

Determining the Optimal Bandwidth Based on Multi-criterion Fusion Proceedngs of 01 4th Internatonal Conference on Machne Learnng and Computng IPCSIT vol. 5 (01) (01) IACSIT Press, Sngapore Determnng the Optmal Bandwdth Based on Mult-crteron Fuson Ha-L Lang 1+, Xan-Mn

More information

Study of Data Stream Clustering Based on Bio-inspired Model

Study of Data Stream Clustering Based on Bio-inspired Model , pp.412-418 http://dx.do.org/10.14257/astl.2014.53.86 Study of Data Stream lusterng Based on Bo-nspred Model Yngme L, Mn L, Jngbo Shao, Gaoyang Wang ollege of omputer Scence and Informaton Engneerng,

More information

Load Balancing for Hex-Cell Interconnection Network

Load Balancing for Hex-Cell Interconnection Network Int. J. Communcatons, Network and System Scences,,, - Publshed Onlne Aprl n ScRes. http://www.scrp.org/journal/jcns http://dx.do.org/./jcns.. Load Balancng for Hex-Cell Interconnecton Network Saher Manaseer,

More information

An Entropy-Based Approach to Integrated Information Needs Assessment

An Entropy-Based Approach to Integrated Information Needs Assessment Dstrbuton Statement A: Approved for publc release; dstrbuton s unlmted. An Entropy-Based Approach to ntegrated nformaton Needs Assessment June 8, 2004 Wllam J. Farrell Lockheed Martn Advanced Technology

More information

For instance, ; the five basic number-sets are increasingly more n A B & B A A = B (1)

For instance, ; the five basic number-sets are increasingly more n A B & B A A = B (1) Secton 1.2 Subsets and the Boolean operatons on sets If every element of the set A s an element of the set B, we say that A s a subset of B, or that A s contaned n B, or that B contans A, and we wrte A

More information

FINDING IMPORTANT NODES IN SOCIAL NETWORKS BASED ON MODIFIED PAGERANK

FINDING IMPORTANT NODES IN SOCIAL NETWORKS BASED ON MODIFIED PAGERANK FINDING IMPORTANT NODES IN SOCIAL NETWORKS BASED ON MODIFIED PAGERANK L-qng Qu, Yong-quan Lang 2, Jng-Chen 3, 2 College of Informaton Scence and Technology, Shandong Unversty of Scence and Technology,

More information

X- Chart Using ANOM Approach

X- Chart Using ANOM Approach ISSN 1684-8403 Journal of Statstcs Volume 17, 010, pp. 3-3 Abstract X- Chart Usng ANOM Approach Gullapall Chakravarth 1 and Chaluvad Venkateswara Rao Control lmts for ndvdual measurements (X) chart are

More information

A Fast Content-Based Multimedia Retrieval Technique Using Compressed Data

A Fast Content-Based Multimedia Retrieval Technique Using Compressed Data A Fast Content-Based Multmeda Retreval Technque Usng Compressed Data Borko Furht and Pornvt Saksobhavvat NSF Multmeda Laboratory Florda Atlantc Unversty, Boca Raton, Florda 3343 ABSTRACT In ths paper,

More information

Collaboratively Regularized Nearest Points for Set Based Recognition

Collaboratively Regularized Nearest Points for Set Based Recognition Academc Center for Computng and Meda Studes, Kyoto Unversty Collaboratvely Regularzed Nearest Ponts for Set Based Recognton Yang Wu, Mchhko Mnoh, Masayuk Mukunok Kyoto Unversty 9/1/013 BMVC 013 @ Brstol,

More information

Concurrent Apriori Data Mining Algorithms

Concurrent Apriori Data Mining Algorithms Concurrent Apror Data Mnng Algorthms Vassl Halatchev Department of Electrcal Engneerng and Computer Scence York Unversty, Toronto October 8, 2015 Outlne Why t s mportant Introducton to Assocaton Rule Mnng

More information

Sum of Linear and Fractional Multiobjective Programming Problem under Fuzzy Rules Constraints

Sum of Linear and Fractional Multiobjective Programming Problem under Fuzzy Rules Constraints Australan Journal of Basc and Appled Scences, 2(4): 1204-1208, 2008 ISSN 1991-8178 Sum of Lnear and Fractonal Multobjectve Programmng Problem under Fuzzy Rules Constrants 1 2 Sanjay Jan and Kalash Lachhwan

More information

From Comparing Clusterings to Combining Clusterings

From Comparing Clusterings to Combining Clusterings Proceedngs of the Twenty-Thrd AAAI Conference on Artfcal Intellgence (008 From Comparng Clusterngs to Combnng Clusterngs Zhwu Lu and Yuxn Peng and Janguo Xao Insttute of Computer Scence and Technology,

More information

Analysis of Collaborative Distributed Admission Control in x Networks

Analysis of Collaborative Distributed Admission Control in x Networks 1 Analyss of Collaboratve Dstrbuted Admsson Control n 82.11x Networks Thnh Nguyen, Member, IEEE, Ken Nguyen, Member, IEEE, Lnha He, Member, IEEE, Abstract Wth the recent surge of wreless home networks,

More information

S1 Note. Basis functions.

S1 Note. Basis functions. S1 Note. Bass functons. Contents Types of bass functons...1 The Fourer bass...2 B-splne bass...3 Power and type I error rates wth dfferent numbers of bass functons...4 Table S1. Smulaton results of type

More information

Clustering Algorithm of Similarity Segmentation based on Point Sorting

Clustering Algorithm of Similarity Segmentation based on Point Sorting Internatonal onference on Logstcs Engneerng, Management and omputer Scence (LEMS 2015) lusterng Algorthm of Smlarty Segmentaton based on Pont Sortng Hanbng L, Yan Wang*, Lan Huang, Mngda L, Yng Sun, Hanyuan

More information

NUMERICAL SOLVING OPTIMAL CONTROL PROBLEMS BY THE METHOD OF VARIATIONS

NUMERICAL SOLVING OPTIMAL CONTROL PROBLEMS BY THE METHOD OF VARIATIONS ARPN Journal of Engneerng and Appled Scences 006-017 Asan Research Publshng Network (ARPN). All rghts reserved. NUMERICAL SOLVING OPTIMAL CONTROL PROBLEMS BY THE METHOD OF VARIATIONS Igor Grgoryev, Svetlana

More information

Tsinghua University at TAC 2009: Summarizing Multi-documents by Information Distance

Tsinghua University at TAC 2009: Summarizing Multi-documents by Information Distance Tsnghua Unversty at TAC 2009: Summarzng Mult-documents by Informaton Dstance Chong Long, Mnle Huang, Xaoyan Zhu State Key Laboratory of Intellgent Technology and Systems, Tsnghua Natonal Laboratory for

More information

Advanced Computer Networks

Advanced Computer Networks Char of Network Archtectures and Servces Department of Informatcs Techncal Unversty of Munch Note: Durng the attendance check a stcker contanng a unque QR code wll be put on ths exam. Ths QR code contans

More information

A Unified Framework for Semantics and Feature Based Relevance Feedback in Image Retrieval Systems

A Unified Framework for Semantics and Feature Based Relevance Feedback in Image Retrieval Systems A Unfed Framework for Semantcs and Feature Based Relevance Feedback n Image Retreval Systems Ye Lu *, Chunhu Hu 2, Xngquan Zhu 3*, HongJang Zhang 2, Qang Yang * School of Computng Scence Smon Fraser Unversty

More information

Classifying Acoustic Transient Signals Using Artificial Intelligence

Classifying Acoustic Transient Signals Using Artificial Intelligence Classfyng Acoustc Transent Sgnals Usng Artfcal Intellgence Steve Sutton, Unversty of North Carolna At Wlmngton (suttons@charter.net) Greg Huff, Unversty of North Carolna At Wlmngton (jgh7476@uncwl.edu)

More information

Complex Numbers. Now we also saw that if a and b were both positive then ab = a b. For a second let s forget that restriction and do the following.

Complex Numbers. Now we also saw that if a and b were both positive then ab = a b. For a second let s forget that restriction and do the following. Complex Numbers The last topc n ths secton s not really related to most of what we ve done n ths chapter, although t s somewhat related to the radcals secton as we wll see. We also won t need the materal

More information

The Greedy Method. Outline and Reading. Change Money Problem. Greedy Algorithms. Applications of the Greedy Strategy. The Greedy Method Technique

The Greedy Method. Outline and Reading. Change Money Problem. Greedy Algorithms. Applications of the Greedy Strategy. The Greedy Method Technique //00 :0 AM Outlne and Readng The Greedy Method The Greedy Method Technque (secton.) Fractonal Knapsack Problem (secton..) Task Schedulng (secton..) Mnmum Spannng Trees (secton.) Change Money Problem Greedy

More information

An Indian Journal FULL PAPER ABSTRACT KEYWORDS. Trade Science Inc.

An Indian Journal FULL PAPER ABSTRACT KEYWORDS. Trade Science Inc. [Type text] [Type text] [Type text] ISSN : 97-735 Volume Issue 9 BoTechnology An Indan Journal FULL PAPER BTAIJ, (9), [333-3] Matlab mult-dmensonal model-based - 3 Chnese football assocaton super league

More information

Support Vector Machines

Support Vector Machines /9/207 MIST.6060 Busness Intellgence and Data Mnng What are Support Vector Machnes? Support Vector Machnes Support Vector Machnes (SVMs) are supervsed learnng technques that analyze data and recognze patterns.

More information

Solving two-person zero-sum game by Matlab

Solving two-person zero-sum game by Matlab Appled Mechancs and Materals Onlne: 2011-02-02 ISSN: 1662-7482, Vols. 50-51, pp 262-265 do:10.4028/www.scentfc.net/amm.50-51.262 2011 Trans Tech Publcatons, Swtzerland Solvng two-person zero-sum game by

More information

A Post Randomization Framework for Privacy-Preserving Bayesian. Network Parameter Learning

A Post Randomization Framework for Privacy-Preserving Bayesian. Network Parameter Learning A Post Randomzaton Framework for Prvacy-Preservng Bayesan Network Parameter Learnng JIANJIE MA K.SIVAKUMAR School Electrcal Engneerng and Computer Scence, Washngton State Unversty Pullman, WA. 9964-75

More information

Parallel matrix-vector multiplication

Parallel matrix-vector multiplication Appendx A Parallel matrx-vector multplcaton The reduced transton matrx of the three-dmensonal cage model for gel electrophoress, descrbed n secton 3.2, becomes excessvely large for polymer lengths more

More information

TPL-Aware Displacement-driven Detailed Placement Refinement with Coloring Constraints

TPL-Aware Displacement-driven Detailed Placement Refinement with Coloring Constraints TPL-ware Dsplacement-drven Detaled Placement Refnement wth Colorng Constrants Tao Ln Iowa State Unversty tln@astate.edu Chrs Chu Iowa State Unversty cnchu@astate.edu BSTRCT To mnmze the effect of process

More information

Unsupervised Learning

Unsupervised Learning Pattern Recognton Lecture 8 Outlne Introducton Unsupervsed Learnng Parametrc VS Non-Parametrc Approach Mxture of Denstes Maxmum-Lkelhood Estmates Clusterng Prof. Danel Yeung School of Computer Scence and

More information

A MOVING MESH APPROACH FOR SIMULATION BUDGET ALLOCATION ON CONTINUOUS DOMAINS

A MOVING MESH APPROACH FOR SIMULATION BUDGET ALLOCATION ON CONTINUOUS DOMAINS Proceedngs of the Wnter Smulaton Conference M E Kuhl, N M Steger, F B Armstrong, and J A Jones, eds A MOVING MESH APPROACH FOR SIMULATION BUDGET ALLOCATION ON CONTINUOUS DOMAINS Mark W Brantley Chun-Hung

More information

The Shortest Path of Touring Lines given in the Plane

The Shortest Path of Touring Lines given in the Plane Send Orders for Reprnts to reprnts@benthamscence.ae 262 The Open Cybernetcs & Systemcs Journal, 2015, 9, 262-267 The Shortest Path of Tourng Lnes gven n the Plane Open Access Ljuan Wang 1,2, Dandan He

More information

Learning-Based Top-N Selection Query Evaluation over Relational Databases

Learning-Based Top-N Selection Query Evaluation over Relational Databases Learnng-Based Top-N Selecton Query Evaluaton over Relatonal Databases Lang Zhu *, Wey Meng ** * School of Mathematcs and Computer Scence, Hebe Unversty, Baodng, Hebe 071002, Chna, zhu@mal.hbu.edu.cn **

More information

Virtual Machine Migration based on Trust Measurement of Computer Node

Virtual Machine Migration based on Trust Measurement of Computer Node Appled Mechancs and Materals Onlne: 2014-04-04 ISSN: 1662-7482, Vols. 536-537, pp 678-682 do:10.4028/www.scentfc.net/amm.536-537.678 2014 Trans Tech Publcatons, Swtzerland Vrtual Machne Mgraton based on

More information

Pruning Training Corpus to Speedup Text Classification 1

Pruning Training Corpus to Speedup Text Classification 1 Prunng Tranng Corpus to Speedup Text Classfcaton Jhong Guan and Shugeng Zhou School of Computer Scence, Wuhan Unversty, Wuhan, 430079, Chna hguan@wtusm.edu.cn State Key Lab of Software Engneerng, Wuhan

More information

R s s f. m y s. SPH3UW Unit 7.3 Spherical Concave Mirrors Page 1 of 12. Notes

R s s f. m y s. SPH3UW Unit 7.3 Spherical Concave Mirrors Page 1 of 12. Notes SPH3UW Unt 7.3 Sphercal Concave Mrrors Page 1 of 1 Notes Physcs Tool box Concave Mrror If the reflectng surface takes place on the nner surface of the sphercal shape so that the centre of the mrror bulges

More information

Random Kernel Perceptron on ATTiny2313 Microcontroller

Random Kernel Perceptron on ATTiny2313 Microcontroller Random Kernel Perceptron on ATTny233 Mcrocontroller Nemanja Djurc Department of Computer and Informaton Scences, Temple Unversty Phladelpha, PA 922, USA nemanja.djurc@temple.edu Slobodan Vucetc Department

More information

Data Preprocessing Based on Partially Supervised Learning Na Liu1,2, a, Guanglai Gao1,b, Guiping Liu2,c

Data Preprocessing Based on Partially Supervised Learning Na Liu1,2, a, Guanglai Gao1,b, Guiping Liu2,c 6th Internatonal Conference on Informaton Engneerng for Mechancs and Materals (ICIMM 2016) Data Preprocessng Based on Partally Supervsed Learnng Na Lu1,2, a, Guangla Gao1,b, Gupng Lu2,c 1 College of Computer

More information

Load-Balanced Anycast Routing

Load-Balanced Anycast Routing Load-Balanced Anycast Routng Chng-Yu Ln, Jung-Hua Lo, and Sy-Yen Kuo Department of Electrcal Engneerng atonal Tawan Unversty, Tape, Tawan sykuo@cc.ee.ntu.edu.tw Abstract For fault-tolerance and load-balance

More information

Problem Definitions and Evaluation Criteria for Computational Expensive Optimization

Problem Definitions and Evaluation Criteria for Computational Expensive Optimization Problem efntons and Evaluaton Crtera for Computatonal Expensve Optmzaton B. Lu 1, Q. Chen and Q. Zhang 3, J. J. Lang 4, P. N. Suganthan, B. Y. Qu 6 1 epartment of Computng, Glyndwr Unversty, UK Faclty

More information

Wishing you all a Total Quality New Year!

Wishing you all a Total Quality New Year! Total Qualty Management and Sx Sgma Post Graduate Program 214-15 Sesson 4 Vnay Kumar Kalakband Assstant Professor Operatons & Systems Area 1 Wshng you all a Total Qualty New Year! Hope you acheve Sx sgma

More information

Subspace clustering. Clustering. Fundamental to all clustering techniques is the choice of distance measure between data points;

Subspace clustering. Clustering. Fundamental to all clustering techniques is the choice of distance measure between data points; Subspace clusterng Clusterng Fundamental to all clusterng technques s the choce of dstance measure between data ponts; D q ( ) ( ) 2 x x = x x, j k = 1 k jk Squared Eucldean dstance Assumpton: All features

More information

Incremental Learning with Support Vector Machines and Fuzzy Set Theory

Incremental Learning with Support Vector Machines and Fuzzy Set Theory The 25th Workshop on Combnatoral Mathematcs and Computaton Theory Incremental Learnng wth Support Vector Machnes and Fuzzy Set Theory Yu-Mng Chuang 1 and Cha-Hwa Ln 2* 1 Department of Computer Scence and

More information

Reducing Frame Rate for Object Tracking

Reducing Frame Rate for Object Tracking Reducng Frame Rate for Object Trackng Pavel Korshunov 1 and We Tsang Oo 2 1 Natonal Unversty of Sngapore, Sngapore 11977, pavelkor@comp.nus.edu.sg 2 Natonal Unversty of Sngapore, Sngapore 11977, oowt@comp.nus.edu.sg

More information

EECS 730 Introduction to Bioinformatics Sequence Alignment. Luke Huan Electrical Engineering and Computer Science

EECS 730 Introduction to Bioinformatics Sequence Alignment. Luke Huan Electrical Engineering and Computer Science EECS 730 Introducton to Bonformatcs Sequence Algnment Luke Huan Electrcal Engneerng and Computer Scence http://people.eecs.ku.edu/~huan/ HMM Π s a set of states Transton Probabltes a kl Pr( l 1 k Probablty

More information

Keywords - Wep page classification; bag of words model; topic model; hierarchical classification; Support Vector Machines

Keywords - Wep page classification; bag of words model; topic model; hierarchical classification; Support Vector Machines (IJCSIS) Internatonal Journal of Computer Scence and Informaton Securty, Herarchcal Web Page Classfcaton Based on a Topc Model and Neghborng Pages Integraton Wongkot Srura Phayung Meesad Choochart Haruechayasak

More information

SAO: A Stream Index for Answering Linear Optimization Queries

SAO: A Stream Index for Answering Linear Optimization Queries SAO: A Stream Index for Answerng near Optmzaton Queres Gang uo Kun-ung Wu Phlp S. Yu IBM T.J. Watson Research Center {luog, klwu, psyu}@us.bm.com Abstract near optmzaton queres retreve the top-k tuples

More information

Adaptive Load Shedding for Windowed Stream Joins

Adaptive Load Shedding for Windowed Stream Joins Adaptve Load Sheddng for Wndowed Stream Jons Bu gra Gedk College of Computng, GaTech bgedk@cc.gatech.edu Kun-Lung Wu, Phlp Yu T.J. Watson Research, IBM {klwu,psyu}@us.bm.com Lng Lu College of Computng,

More information

Life Tables (Times) Summary. Sample StatFolio: lifetable times.sgp

Life Tables (Times) Summary. Sample StatFolio: lifetable times.sgp Lfe Tables (Tmes) Summary... 1 Data Input... 2 Analyss Summary... 3 Survval Functon... 5 Log Survval Functon... 6 Cumulatve Hazard Functon... 7 Percentles... 7 Group Comparsons... 8 Summary The Lfe Tables

More information

CHAPTER 2 DECOMPOSITION OF GRAPHS

CHAPTER 2 DECOMPOSITION OF GRAPHS CHAPTER DECOMPOSITION OF GRAPHS. INTRODUCTION A graph H s called a Supersubdvson of a graph G f H s obtaned from G by replacng every edge uv of G by a bpartte graph,m (m may vary for each edge by dentfyng

More information

Private Information Retrieval (PIR)

Private Information Retrieval (PIR) 2 Levente Buttyán Problem formulaton Alce wants to obtan nformaton from a database, but she does not want the database to learn whch nformaton she wanted e.g., Alce s an nvestor queryng a stock-market

More information

Edge Detection in Noisy Images Using the Support Vector Machines

Edge Detection in Noisy Images Using the Support Vector Machines Edge Detecton n Nosy Images Usng the Support Vector Machnes Hlaro Gómez-Moreno, Saturnno Maldonado-Bascón, Francsco López-Ferreras Sgnal Theory and Communcatons Department. Unversty of Alcalá Crta. Madrd-Barcelona

More information

MULTISPECTRAL IMAGES CLASSIFICATION BASED ON KLT AND ATR AUTOMATIC TARGET RECOGNITION

MULTISPECTRAL IMAGES CLASSIFICATION BASED ON KLT AND ATR AUTOMATIC TARGET RECOGNITION MULTISPECTRAL IMAGES CLASSIFICATION BASED ON KLT AND ATR AUTOMATIC TARGET RECOGNITION Paulo Quntlano 1 & Antono Santa-Rosa 1 Federal Polce Department, Brasla, Brazl. E-mals: quntlano.pqs@dpf.gov.br and

More information

The Codesign Challenge

The Codesign Challenge ECE 4530 Codesgn Challenge Fall 2007 Hardware/Software Codesgn The Codesgn Challenge Objectves In the codesgn challenge, your task s to accelerate a gven software reference mplementaton as fast as possble.

More information

Parallel Implementation of Classification Algorithms Based on Cloud Computing Environment

Parallel Implementation of Classification Algorithms Based on Cloud Computing Environment TELKOMNIKA, Vol.10, No.5, September 2012, pp. 1087~1092 e-issn: 2087-278X accredted by DGHE (DIKTI), Decree No: 51/Dkt/Kep/2010 1087 Parallel Implementaton of Classfcaton Algorthms Based on Cloud Computng

More information

Node Classification in Dynamic Social Networks

Node Classification in Dynamic Social Networks Node Classfcaton n Dynamc Socal Networks Yulong Pe, Janpeng Zhang, George H.L. Fletcher, and Mykola Pechenzky Department of Mathematcs and Computer Scence Endhoven Unversty of Technology, Endhoven, the

More information

Quality Improvement Algorithm for Tetrahedral Mesh Based on Optimal Delaunay Triangulation

Quality Improvement Algorithm for Tetrahedral Mesh Based on Optimal Delaunay Triangulation Intellgent Informaton Management, 013, 5, 191-195 Publshed Onlne November 013 (http://www.scrp.org/journal/m) http://dx.do.org/10.36/m.013.5601 Qualty Improvement Algorthm for Tetrahedral Mesh Based on

More information

Meta-heuristics for Multidimensional Knapsack Problems

Meta-heuristics for Multidimensional Knapsack Problems 2012 4th Internatonal Conference on Computer Research and Development IPCSIT vol.39 (2012) (2012) IACSIT Press, Sngapore Meta-heurstcs for Multdmensonal Knapsack Problems Zhbao Man + Computer Scence Department,

More information

Adaptive Load Shedding for Windowed Stream Joins

Adaptive Load Shedding for Windowed Stream Joins Adaptve Load Sheddng for Wndowed Stream Jons Buğra Gedk, Kun-Lung Wu, Phlp S. Yu, Lng Lu College of Computng, Georga Tech Atlanta GA 333 {bgedk,lnglu}@cc.gatech.edu IBM T. J. Watson Research Center Yorktown

More information

Optimal Workload-based Weighted Wavelet Synopses

Optimal Workload-based Weighted Wavelet Synopses Optmal Workload-based Weghted Wavelet Synopses Yoss Matas School of Computer Scence Tel Avv Unversty Tel Avv 69978, Israel matas@tau.ac.l Danel Urel School of Computer Scence Tel Avv Unversty Tel Avv 69978,

More information

Improvement of Spatial Resolution Using BlockMatching Based Motion Estimation and Frame. Integration

Improvement of Spatial Resolution Using BlockMatching Based Motion Estimation and Frame. Integration Improvement of Spatal Resoluton Usng BlockMatchng Based Moton Estmaton and Frame Integraton Danya Suga and Takayuk Hamamoto Graduate School of Engneerng, Tokyo Unversty of Scence, 6-3-1, Nuku, Katsuska-ku,

More information

Multiple Information Sources Cooperative Learning

Multiple Information Sources Cooperative Learning Multple Informaton Sources Cooperatve Learnng Xngquan Zhu Faculty of Eng. & Info. Technology Unversty of Technology, Sydney, Australa xqzhu@t.uts.edu.au Ruomng Jn Dept. of Computer Scence Kent State Unversty,

More information

y and the total sum of

y and the total sum of Lnear regresson Testng for non-lnearty In analytcal chemstry, lnear regresson s commonly used n the constructon of calbraton functons requred for analytcal technques such as gas chromatography, atomc absorpton

More information

Video Proxy System for a Large-scale VOD System (DINA)

Video Proxy System for a Large-scale VOD System (DINA) Vdeo Proxy System for a Large-scale VOD System (DINA) KWUN-CHUNG CHAN #, KWOK-WAI CHEUNG *# #Department of Informaton Engneerng *Centre of Innovaton and Technology The Chnese Unversty of Hong Kong SHATIN,

More information

Virtual Memory. Background. No. 10. Virtual Memory: concept. Logical Memory Space (review) Demand Paging(1) Virtual Memory

Virtual Memory. Background. No. 10. Virtual Memory: concept. Logical Memory Space (review) Demand Paging(1) Virtual Memory Background EECS. Operatng System Fundamentals No. Vrtual Memory Prof. Hu Jang Department of Electrcal Engneerng and Computer Scence, York Unversty Memory-management methods normally requres the entre process

More information

A Novel Adaptive Descriptor Algorithm for Ternary Pattern Textures

A Novel Adaptive Descriptor Algorithm for Ternary Pattern Textures A Novel Adaptve Descrptor Algorthm for Ternary Pattern Textures Fahuan Hu 1,2, Guopng Lu 1 *, Zengwen Dong 1 1.School of Mechancal & Electrcal Engneerng, Nanchang Unversty, Nanchang, 330031, Chna; 2. School

More information

A fast algorithm for color image segmentation

A fast algorithm for color image segmentation Unersty of Wollongong Research Onlne Faculty of Informatcs - Papers (Arche) Faculty of Engneerng and Informaton Scences 006 A fast algorthm for color mage segmentaton L. Dong Unersty of Wollongong, lju@uow.edu.au

More information

A Clustering Algorithm for Chinese Adjectives and Nouns 1

A Clustering Algorithm for Chinese Adjectives and Nouns 1 Clusterng lgorthm for Chnese dectves and ouns Yang Wen, Chunfa Yuan, Changnng Huang 2 State Key aboratory of Intellgent Technology and System Deptartment of Computer Scence & Technology, Tsnghua Unversty,

More information

Proper Choice of Data Used for the Estimation of Datum Transformation Parameters

Proper Choice of Data Used for the Estimation of Datum Transformation Parameters Proper Choce of Data Used for the Estmaton of Datum Transformaton Parameters Hakan S. KUTOGLU, Turkey Key words: Coordnate systems; transformaton; estmaton, relablty. SUMMARY Advances n technologes and

More information

Machine Learning. Topic 6: Clustering

Machine Learning. Topic 6: Clustering Machne Learnng Topc 6: lusterng lusterng Groupng data nto (hopefully useful) sets. Thngs on the left Thngs on the rght Applcatons of lusterng Hypothess Generaton lusters mght suggest natural groups. Hypothess

More information

Array transposition in CUDA shared memory

Array transposition in CUDA shared memory Array transposton n CUDA shared memory Mke Gles February 19, 2014 Abstract Ths short note s nspred by some code wrtten by Jeremy Appleyard for the transposton of data through shared memory. I had some

More information

Cordial and 3-Equitable Labeling for Some Star Related Graphs

Cordial and 3-Equitable Labeling for Some Star Related Graphs Internatonal Mathematcal Forum, 4, 009, no. 31, 1543-1553 Cordal and 3-Equtable Labelng for Some Star Related Graphs S. K. Vadya Department of Mathematcs, Saurashtra Unversty Rajkot - 360005, Gujarat,

More information