Accumulated-Recognition-Rate Normalization for Combining Multiple On/Off-Line Japanese Character Classifiers Tested on a Large Database

Size: px
Start display at page:

Download "Accumulated-Recognition-Rate Normalization for Combining Multiple On/Off-Line Japanese Character Classifiers Tested on a Large Database"

Transcription

1 4 th Internatonal Workshop on Multple Classfer Systems (MCS23) Guldford, UK Accumulated-Recognton-Rate Normalzaton for Combnng Multple On/Off-Lne Japanese Character Classfers Tested on a Large Database Ondrej Velek, Stefan Jaeger, Masak Nakagawa Graduate School of Technology, Tokyo Unversty of Agrculture and Technology Naka-cho Kogane-sh, Tokyo, , Japan E-mal: velek@hands.e.tuat.ac.jp, stefan@hands.e.tuat.ac.jp, nakagawa@cc.tuat.ac.jp Abstract. Ths paper presents a technque for normalzng lkelhood of multple classfers, allowng ther far combnaton. Our technque generates for each recognzer one general or several stroke-number specfc characterstc functons. A smple warpng process maps output scores nto an deal characterstc. A novelty of our approach s n usng a characterstc based on the accumulated recognton rate, whch makes our method very robust and stable to random errors n tranng data and requres no smoothng pror to normalzaton. In ths paper we test our method on a large database named Kuchbue_d, a publcly avalable benchmark for on-lne Japanese handwrtten character recognton and very often used for benchmarkng new methods. 1. Introducton Combnng dfferent classfers for the same classfcaton problem has become very popular durng the last years [7,9]. In handwrtng recognton, classfer combnatons are of partcular nterest snce they allow brdgng the gap between on-lne and offlne handwrtng recognton. An ntegrated on-lne/off-lne recognton system can explot valuable on-lne nformaton whle off-lne data guarantees robustness aganst stroke order and stroke number varatons. Snce the dfferent nature of on-lne and off-lne data complcates ther combnaton, most approaches combne both types of nformaton ether durng pre-processng;.e. feature computaton [1-3] or, lke ths paper, n post-processng [5-8]. In ths paper, we report experments on the combnaton of on-lne and off-lne recognzers for on-lne recognton of Japanese characters. Multple classfers, especally on-lne and off-lne classfers, very often generate lkelhood values that are ncompatble. The focus of ths paper les on our accumulated-recognton-rate normalzaton technque, whch tres to overcome ths problem by algnng the lkelhood values wth the actual performance of each classfer. At frst we defne the accumulated recognton rate, whch s normalzed to a lnearly growng functon. Usng normalzed lkelhood ensures a far combnaton of classfers. We have ntroduced our warpng technque n [13, 16], and tested ts effcency on a small NTT-AT database. That database s a collecton of patterns wrtten by elderly people, usually by ncorrect wrtng style. We have acheved mprovement of the recognton rate from 89.7% for the best sngle classfer to 94.14% by combnng seven classfers n our combnaton system [13] and n [16] 93.68% by combnng two classfers. However, snce the NTT-AT database s not a collecton of typcal 196

2 4 th Internatonal Workshop on Multple Classfer Systems (MCS23) Guldford, UK character patterns and the number of samples and categores s low, a good performance cannot be easly generalzed for common Japanese character patterns. In ths paper we experment wth the benchmark database Kuchbue_d, whch contans onlne Japanese handwrtten characters for whch many researchers have already publshed recognton rates. Ths paper s structured as follows: Secton 2 and 3 outlne the work descrbed n our prevous paper [13] and descrbe the general theoretcal framework of our normalzaton technque based on accumulated recognton rate. In Secton 4 we ntroduce our mproved approach utlzng nformaton about stroke number, frstly publshed n [16]. Secton 5 descrbes sngle classfers and combnaton schemes used n our experments. Secton 6 presents our new results on the Kuchbue_d benchmark set, whch allows comparson of our recognton results on the same benchmark wth other research groups or wth our prevous results on the NTT-AT database (Secton 7). Fnally, Secton 8 concludes ths paper wth a general dscusson of our results. 2. Comparablty of dfferent classfers Let A be a classfer that maps an unknown nput pattern x to one of m possble classes (ω 1,, ω m ), and returns values a = A(x, ω ) denotng ts confdence that x s a member of class ω,. For an deal classfer, each returned value a corresponds to the true probablty of ω gven x: P (ω, x), also called a-posteror probablty, wth a =P (ω, x) 1. In real practce, however, the output values a can merely be approxmatons of the correct a-posteror probabltes. Many classfers do not even output approxmatons but only values related to the a-posteror probabltes. For nstance, A(x, ω ) = a very often denotes the dstance between the nput pattern x and the class ω n a hgh-dmensonal feature space under a gven metrc, sometmes wth a [; ]. In ths case, the best canddate s not the class wth the maxmum value among all a, but the class havng the smallest a ;.e., the class wth the shortest dstance. Also, the scale of output values s generally unknown. For example, from a r = 2a p one cannot predct that class ω r s twce as lkely as class ω p. These nadequaces generally pose lttle problem for a sngle classfer that needs to fnd only k-best canddates. If a sngle classfer rejects patterns wth confdence values below a certan threshold, a relaton between confdence and a-posteror probablty s useful for settng the threshold. However, for combnng several classfers {A, B, C } we necessarly need some relaton among canddates stemmng from dfferent classfers {a =A(x, ω ), b =B(x, ω ), } n order to compare them and to select the best class. To better descrbe the output of classfers, we defne two auxlary functons, n(a k ) and n correct (a k ) countng the overall number of samples and correctly recognzed samples for each output value respectvely: Functon n(a k ) returns the number of test patterns classfed wth output value a k and functon n correct (a k ) returns the number of correctly recognzed test patterns. We begn by countng the number of correctly and ncorrectly recognzed samples for each value: n correct (a k ) and n ncorrect (a k )= n t (a k )- n correct (a k ). Graph 1 and Graph 2 show two exemplary hstograms of these numbers for off-lne and on-lne recognton respectvely. An deal classfer returns only correct answers wth lkelhood values coverng the whole range of possble values. As a matter of fact, most practcal classfers also 197

3 4 th Internatonal Workshop on Multple Classfer Systems (MCS23) Guldford, UK return ncorrect answers and do not cover all output values. Graph 1 and Graph 2 llustrate ths for the off-lne classfer - most of the correct output s around a =6 -, whle the on-lne classfer has most of ts correct answers around a=9. The peak of n correct (a) dffers from n ncorrect () n both classfers. Also, both classfers use only a small range of the output nterval [;1] ntensvely. 25 Hstogram of a confdence value a 5 Hstogram of a confdence value a [a ] correct ncorrect [a] correct ncorrect Graph 1: Hstogram for an off-lne recognzer Graph 2: Hstogram for an on-lne recognzer. n correct For a sngle classfer system, only the recognton rate;.e., the number of correctly recognzed patterns () dvded by the number of overall patterns, s of mportance. However, for combnng multple classfers, not only the recognton rate, but also the dstrbuton of n correct ( a k ) and n ncorrect ( a k ) s of nterest. If some classfers provde better recognton rates than the best sngle recognton rate on several sub-ntervals, then we can suppose that by combnng multple classfers the combned recognton rate wll outperform the sngle best rate. Graph 3 and Graph 4 show the recognton rates correspondng to Graph 1 and 2 respectvely. Recognton Rate R(a) [% ] correct nccorect [a ] Recognton Rate R(a) [% ] correct nccorect [a ] Graph 3: Rcg.Rate for an off-lne recognzer. Graph 4: Rcg.Rate for an on-lne recognzer. 3. Normalzaton and Warpng Our goal s to normalze the output of each classfer so that t better reflects the actual classfer performance for each output value and allows drect comparson and combnaton wth outputs from other classfers. The man dea s to turn confdence values nto a monotone ncreasng functon dependng on the recognton rate. After normalzaton, the hghest output should stll defne the most lkely class of an unknown nput pattern x; a > a j should mply that P(ω, x) > P(ω j x). However straghtforward normalzaton of classfer outputs to a monotone ncreasng functon s not possble, because classfer confdence s nether a contnuous nor a monotone functon. 198

4 4 th Internatonal Workshop on Multple Classfer Systems (MCS23) Guldford, UK In our new approach we solve ths problem by defnng the accumulated recognton rate (whch s always a contnuous functon) and applyng a warpng process on t. Ths process expands and compresses local subntervals of the horzontal lkelhood-axs to better match the practcal recognton rates acheved, wthout changng the order and wthout addng artfcal error by addtve smoothng. The confdence value s often ether smaller than the value suggested by the recognton rate, whch means that the confdence value s too pessmstc, or t s hgher than the actual recognton rate suggests, whch s then a too optmstc value. We proposed usng a second tranng set to measure the classfer performance for each lkelhood value, gven an approprate quantzaton of lkelhood, and then calbrate lkelhood n a post-processng step accordng to the performance measured. Let A={a,, a,, a max } be a set of lkelhood values wth a beng the lowest and a max beng the hghest lkelhood assgned by the classfer. In our experments, the lkelhood values span the nteger nterval rangng from to 1, and thus A = [;1], max = 1, and a k = k. The calbraton replaces the old lkelhood new values a by ther correspondng recognton rates r = a so that after normalzaton lkelhood and recognton rate are equal: A =R Usng our auxlary functons, we can state ths as follows: new ncorrect ( a ) amax a =. n( a ) Graph 5 and Graph 6 show the accumulated recognton rates before and after normalzaton computed for the off-lne and on-lne recognton rates shown n Graph 3 and Graph 4 respectvely. The error rates depcted n Graph 5 and Graph 6 show the remanng percentage of msclassfed patterns for lkelhood values hgher than a partcular value a. [%].9 Accumulated Recognton Rate.8 Accumulated RR(,1)=Rec.Rate [a ] correct (before) ncorrect (before) correct (after) ncorrect (after) [%].9 Accumulated Recognton Rate.8 Accumulated RR(,1)=Rec.Rate [a ] correct (before) ncorrect (before) correct (after) ncorrect (after) Graph 5: R(<,a >) Accumulated Recognton Rate for an off-lne recognzer before and after normalzaton. Graph 6: R(<,a >) Accumulated Recognton Rate for an on-lne recognzer before and after normalzaton. Note that the accumulated recognton rate s a monotone growng functon over classfer output. Our normalzaton method equals the accumulated lkelhood mass wth the correspondng accumulated recognton rate. Usng the nomenclature ntroduced 199

5 4 th Internatonal Workshop on Multple Classfer Systems (MCS23) Guldford, UK above, our normalzaton ensures the followng equaton: as A R =, whch translates 1 1 n( a ) a = n ( a ) for all, wth N beng the overall number of k k correct L k= N k = patterns and L beng the overall lkelhood mass;.e., k max = L n( a k )* a k. For each classfer, we thus normalze the output so that the accumulated probablty functon R(<,a >) becomes a functon proportonal to the classfer output. Accordngly, we adjust each classfer s output by addng the followng adjustment charf(a ): n ( ) k correct a k a = + amax = a = a + charf ( a ) N n ( ) ( ) k correct a k n k correct a = k a ' amax a = = = + amax a = a + charf ( a ) N N where a max s the maxmum possble output of a classfer ( a max = 1 n our experments), N s the number of overall patterns, and r stands for the partally accumulated recognton rate. We call ths classfer-specfc adjustment the characterstc functon [charf ] of a classfer. To compute the [charf ] of a classfer, we need another sample set, whch should be ndependent from the tranng set. Hence, we use data ndependent from the tranng set to ensure a proper evaluaton, although we observed that a characterstc functon depends mostly on the classfer and not the data. To llustrate the effects of normalzaton we now compare the graphs 1-4 wth the correspondng graphs computed after normalzaton;.e., after addng the adjustment. Graph 7 and Graph 8 ncely llustrate the unform dstrbuton of output values after normalzaton, compared to Graph 1 and Graph 2. In both Graphs 7-8, we see that, after normalzaton, the whole output spectrum s used for correct answers. Moreover, ncorrect answers concentrate n the left, low-value part wth a peak near zero. Snce both off-lne and on-lne recognzers show the same output behavor after normalzaton, ths provdes us wth a standard measure for classfer combnaton. k= Hstogram of a confdence value a Hstogram of a confdence value a correct ncorrect [a ] correct ncorrect [a ] Graph 7: Hstogram for an off-lne recognzer after normalzaton. Graph 8: Hstogram for an on-lne recognzer after normalzaton 2

6 4 th Internatonal Workshop on Multple Classfer Systems (MCS23) Guldford, UK Fnally, Graph 9 and Graph 1 show the lkelhood-dependent recognton rates for each lkelhood value after normalzaton, whch correspond to the recognton rates shown n Graph 3 and Graph 4 before normalzaton. [% ] Recognton Rate R(a) correct nccorect [a ] [% ] Recognton Rate R(a) correct nccorect [a ] Graph 9: Recognton Rate for an off-lne recognzer after normalzaton. Graph 1: Recognton Rate for an on-lne recognzer after normalzaton In Graph 9 and Graph 1, the recognton performance quckly ncreases to a hgh level, and except for very small lkelhood values, the recognton rate s about the same for each lkelhood value. The number of correctly recognzed patterns s hgher than the number of falsely recognzed ones, from lkelhood values greater than 5 onwards. Agan, both off-lne and on-lne classfers behave smlarly here. 4. Characterstc of On/Off classfers accordng to stroke number Number of strokes s the basc feature of Chnese and Japanese characters. The rght stroke number can be found n a dctonary and vares from 1 to about 3. However, for fluently and quckly wrtten characters, the number of strokes s often lower because some strokes are connected. In some cases the number of strokes can be hgher than they should be. An nterestng characterstc of stroke number varatons s gven n [11]. Graph show the recognton rate accordng to the stroke number for an on-lne and off-lne recognzer respectvely. Even from a bref vew, we see the bg dfference between both classfers. An off-lne recognzer s weaker n recognzng characters wrtten wth a low number of strokes. The recognton rate grows wth ncreasng complexty of patterns. Ths s n accordance wth the well-known fact that for off-lne classfers t s more dffcult to recognze smple patterns lke Kana than dffcult Kanj. On the contrary, on-lne recognzers are very effcent for one-stroke patterns. 21

7 4 th Internatonal Workshop on Multple Classfer Systems (MCS23) Guldford, UK [%] Recognton rate for off-lne recognzer [%] Recognton rate for on-lne recognzer number of strokes number of strokes Graph 11: Recognton rate accordng to the stroke number for off-lne recognzer. Graph 12: Recognton rate accordng to the stroke number for on-lne recognzer. From Graph 12 we see that although an average rate of the on-lne recognzer s about 5% worse than that of the off-lne recognzer, for characters wrtten by one, two, or three strokes the recognton rate s better, or at least smlar. Snce classfers effcency depends on the strokes number, we tred to make not only one general characterstc functon, but 14 specfc functons, where the last one s for patterns wrtten by fourteen or more strokes. In Graph 13 we show some examples of stroke-dependent characterstc functons [charf ] for an on-lne recognzer and n Graph 14 for an offlne recognzer. [char ] #1 #14 Characterstc functons [char ] #5 # #1 number of strokes [a ] # 1 # 2 # 5 # 1 # 14 Graph 13: Characterstc functons for onlne recognzer [char ] Characterstc functons [char ] #14 # #5-2 #2 #1 [a number of strokes ] # 1 # 2 # 5 # 1 # 14 Graph 14: Characterstc functons for offlne recognzer 5. On-lne & off-lne classfers and combnaton schemes Our off-lne recognzer represents each character as a 256-dmensonal feature vector. Every nput pattern s scaled to a 64x64 grd by non-lnear normalzaton and smoothed by a connectvty-preservng procedure. Then, a normalzed mage s decomposed nto 4 contour sub-patterns, one for each man orentaton. Fnally, a 64- dmensonal feature vector s extracted for each contour pattern from ts convoluton wth a blurrng mask (Gaussan flter). A pre-classfcaton step precedes the actual fnal recognton. Pre-classfcaton selects the 5 canddates wth the shortest Eucldan dstances between the categores mean vectors and the test pattern. The fnal classfcaton uses a modfed quadratc dscrmnant functon (MQDF2) developed by 22

8 4 th Internatonal Workshop on Multple Classfer Systems (MCS23) Guldford, UK Kmura et al. [5] from tradtonal QDF. The off-lne classfer s traned wth the offlne HP-JEITA database (3,36 categores, 58 patterns per category). The on-lne recognzer presented n [12] by Nakagawa et al. employs a two-step coarse classfcaton based on easly extractable on-lne features n the frst step and four-drectonal features n the second step. An effcent elastc matcher explotng herarchcal pattern representatons performs the man classfcaton. The recognzer s traned wth the on-lne Nakayos_t database[6] (163 patterns for each of 4,438 categores). We nvestgate two dfferent combnaton strateges for combnng our on-lne and off-lne recognzers: max-rule and sum-rule. Max-rule takes the class wth the maxmum output value among each classfer, whle the sum-rule adds up the output for each class and selects the one wth the hghest sum[7]. 6. Benchmarkng wth the Kuchbue_d database Ths secton presents experments evaluatng our proposed normalzaton wth respect to the Kuchbue_d database [6], [14], whch s a wdely acknowledged benchmark for Japanese character recognton. It contans handwrtten sentences from a Japanese newspaper. In total, Kuchbue contans more than 1.4 mllon characters wrtten by 12 wrters (11,962 Kanj samples per wrter). Kuchbue covers 3,356 Kanj categores ncludng the two phonetc alphabets Hragana and Katakana, plus alphanumercal characters. Snce our recently publshed results were based on the ETL9B benchmark-database, whch does not cover Katakana and alphanumercal characters, we confne our experments to the 336 categores of Kuchbue (Kanj and Hragana) ncluded n ETL9B. On-lne patterns processed by the off-lne classfer are converted to btmaps by our callgraphc method [15] pantng realstc off-lne patterns. Table 1: Combnaton of an on-lne and an off-lne recognzer, tested on Kuchbue_d. Benchmark: Kuchbue_d wthout normalzaton Normalzaton by 1 characterstc functon Normalzaton by 14 characterstc functon Sngle classfer Combnaton of sngle classfers On-lne Off-lne AND OR Max Sum The effcency of our two sngle recognzers are 91.6% (on-lne) and 94.69% (offlne); Table 1, column 1 and 2. The hgher recognton rate of the off-lne recognzer s partally result of more tranng patterns per category (58 aganst 163) and smaller scope of recognzable categores (336 aganst 4438). The next two columns show the theoretcal worst (AND - a pattern was recognzed by both classfers) and best (OR - a pattern was recognzed at least by one recognzer) combnaton schemes, whch are the theoretcal bounds for any combnaton rule. Especally the OR combnaton scheme s mportant. Although t can never be utlzed n a real applcaton, the best theoretcal boundary s mportant for comparng the effcency of other combnaton schemes. 23

9 4 th Internatonal Workshop on Multple Classfer Systems (MCS23) Guldford, UK Column 5 and Column 6 of Table 1 show the recognton rates for the max-rule and sum-rule respectvely. The sum-rule performs better than max-rule. A theoretcal explanaton why the sum-rule outperforms the max-rule s gven n [7]. The sum-rule s also more resstant to problems of non-normalzed lkelhood. We have acheved mprovements of sngle recognton rates from 91.6% and 94.69% up to 97.71%. And what s also mportant, our result s very near to the theoretcal maxmum 98.88% of the OR combnaton scheme. 7. Benchmarkng wth the NTT-AT database In ths secton we compare the results based on Kuchbue_d from Secton 6 wth results based on the NTT-AT database publshed n [16]. The NTT-AT database contans data wrtten by elderly people wth an average age of 7.5 years, wth the oldest wrter beng 86 years old. These patterns are casually wrtten, very often wth an untypcal stroke order. Table 2: Combnaton of on-lne and off-lne recognzers tested on NTT-AT. Benchmark: NTT-AT wthout normalzaton Sngle classfer Combnaton of sngle classfers On-lne Off-lne AND OR Max Sum Normalzaton by 1 characterstc functon Normalzaton by 14 characterstc functon Table 2 shows that our normalzaton sgnfcantly mproves the recognton rate. From 85.67% (on-lne classfer) and 89.7% (off-lne classfer) to 93.68%. 8. Dscusson In ths paper we presented an updated verson of our accumulated-recognton-ratebased normalzaton for combnng multple classfers, whch was ntroduced at IWFHR 21 [13,16]. A smple warpng process algns confdence values accordng to the accumulated recognton rate, so that the normalzed values allow a far combnaton. In our prevous experments on the NTT-AT database, whch contans casually wrtten characters often wrtten n untypcal wrtng order, we mproved the recognton rate from 89.7% for the best sngle recognzer to 93.68%. In ths paper we used the Kuchbue_d database, whch s the well-establshed benchmark of common casually wrtten Japanese characters. Its sze s 23 tmes bgger than that of the NTT-AT database. The mprovement of the recognton rate s from 94.69% for the best sngle recognzer to 96.29% for the sum-rule combnaton pror to our normalzaton, 97.13% when normalzed by one common characterstc functon, and fnally 97.71% f normalzed by 14 stroke-number-specfc characterstc functons. The man advantages of our normalzaton based on the accumulated-recogntonrate are as follows: Frstly, the possblty to combne any number of classfers, wthout havng weaker classfers degrade the fnal result. Secondly, t allows easy comb- 24

10 4 th Internatonal Workshop on Multple Classfer Systems (MCS23) Guldford, UK naton of classfers wth ncompatble confdence values. Thrdly, normalzng s a completely automatc process, wthout emprcal parameter settngs. And fourthly, normalzaton s performed only once and does not consume tme durng recognton. Normalzaton based on stroke numbers better reflects the nature of each recognzer and thus leads to better results; t requres more tranng patterns for computng multple characterstc functons though. Although our accumulated-recognton-rate-based normalzaton was tested on combned on-lne/off-lne classfers for on-lne Japanese handwrtten characters, t should be useful for combnatons n varous felds of pattern recognton. References 1. M. Hamanaka, K. Yamada, J. Tsukumo, On-Lne Japanese Character Recognton Experments by an Off-Lne Method Based on Normalzaton-Cooperated Feature Extracton, Proc. 2 nd ICDAR (1993) M. Okamoto, A. Nakamura, K. Yamamoto, Drecton-Change Features of Imagnary Strokes for On-Lne Handwrtng Character Recognton, 14 th ICPR (1998) S. Jaeger, S. Manke, J. Rechert, A. Wabel, On-Lne Handwrtng Recognton: The Npen++ Recognzer, IJDAR 3(3) (21) F. Kmura, et al., Modfed quadratc dscrmnant functon and the applcaton to Chnese characters, IEEE Pattern Analyss and Machne Intellgence, 9(1), pp , H. Kang, K. Km and J. Km, A Framework for Probablstc Combnaton of Multple Classfers at an Abstract Level, EAAI 1 (4) (1997) Dstrbuton of the on-lne databases Kuchbue_d and Nakayos_t, Nakagawa laboratory, TUAT, Japan: 7. J.Kttler, M.Hatef, R.Dun, J.Matas, On Combnng Classfers, IEEE PAMI 2(3) (1998) H. Tanaka, K. Nakajma, K. Ishgak, K. Akyama, M. Nakagawa, Hybrd Pen-Input Character Recognton System Based on Integraton of On-Lne Off-Lne Recognton, Proc. 5 th ICDAR (1999) T. Ho, J. Hull, S. Srhar, Decson Combnaton n Multple Classfer Systems, IEEE Trans. on Pattern Analyss and Machne Intellgence, vol. 16, no. 1, pp , M. Oberländer, German Patent DE C1 (n German), 1995, on-lne avalable on the webste of the German Patent and Trade Mark Offce: K. Matsumoto, T. Fukushma, M. Nakagawa, Collecton and analyss of on-lne handwrtten Japanese character patterns, Proc. 6 th ICDAR, Seattle, 21, pp M. Nakagawa, K. Akyama, L.V. Tu, A. Homma, T. Hgashyama, Robust and Hghly Customzable Recognton of On-Lne Handwrtten Japanese Characters, Proc. 13 th ICPR (1996), volume III, O. Velek, S. Jaeger, M. Nakagawa, A New Warpng Technque for Normalzng Lkelhood of Multple Classfers and ts Effectveness n Combned On-Lne/Off-Lne Japanese Character Recognton, Proc. 8 th IWFHR(22), pp S. Jaeger, M. Nakagawa, Two On-Lne Japanese Character Databases n Unpen Format, Proc. 6 th ICDAR(21), pp O. Velek, M. Nakagawa, C.-L.Lu, Vector-to-Image Transformaton of Character patterns for On-lne and Off-lne Recognton, Internatonal Journal of Computer Processng of Orental Languages, Vol.15, No2 (22) O. Velek, M.Nakagawa, Usng Stroke-Number Characterstcs for Improvng Effcency of Combned On-Lne and Off-Lne Japanese Character Classfers, 5 th DAS (22),

Incremental MQDF Learning for Writer Adaptive Handwriting Recognition 1

Incremental MQDF Learning for Writer Adaptive Handwriting Recognition 1 200 2th Internatonal Conference on Fronters n Handwrtng Recognton Incremental MQDF Learnng for Wrter Adaptve Handwrtng Recognton Ka Dng, Lanwen Jn * School of Electronc and Informaton Engneerng, South

More information

An Approach to Real-Time Recognition of Chinese Handwritten Sentences

An Approach to Real-Time Recognition of Chinese Handwritten Sentences An Approach to Real-Tme Recognton of Chnese Handwrtten Sentences Da-Han Wang, Cheng-Ln Lu Natonal Laboratory of Pattern Recognton, Insttute of Automaton of Chnese Academy of Scences, Bejng 100190, P.R.

More information

th International Conference on Document Analysis and Recognition

th International Conference on Document Analysis and Recognition 2013 12th Internatonal Conference on Document Analyss and Recognton Onlne Handwrtten Cursve Word Recognton Usng Segmentaton-free n Combnaton wth P2DBM-MQDF Blan Zhu 1, Art Shvram 2, Srrangaraj Setlur 2,

More information

An Empirical Comparative Study of Online Handwriting Chinese Character Recognition:Simplified v.s.traditional

An Empirical Comparative Study of Online Handwriting Chinese Character Recognition:Simplified v.s.traditional 2013 12th Internatonal Conference on Document Analyss and Recognton An Emprcal Comparatve Study of Onlne Handwrtng Chnese Recognton:Smplfed v.s.tradtonal Yan Gao, Lanwen Jn +, Wexn Yang School of Electronc

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

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

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

Outline. Discriminative classifiers for image recognition. Where in the World? A nearest neighbor recognition example 4/14/2011. CS 376 Lecture 22 1

Outline. Discriminative classifiers for image recognition. Where in the World? A nearest neighbor recognition example 4/14/2011. CS 376 Lecture 22 1 4/14/011 Outlne Dscrmnatve classfers for mage recognton Wednesday, Aprl 13 Krsten Grauman UT-Austn Last tme: wndow-based generc obect detecton basc ppelne face detecton wth boostng as case study Today:

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

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

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

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

Skew Angle Estimation and Correction of Hand Written, Textual and Large areas of Non-Textual Document Images: A Novel Approach

Skew Angle Estimation and Correction of Hand Written, Textual and Large areas of Non-Textual Document Images: A Novel Approach Angle Estmaton and Correcton of Hand Wrtten, Textual and Large areas of Non-Textual Document Images: A Novel Approach D.R.Ramesh Babu Pyush M Kumat Mahesh D Dhannawat PES Insttute of Technology Research

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

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

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

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

FEATURE EXTRACTION. Dr. K.Vijayarekha. Associate Dean School of Electrical and Electronics Engineering SASTRA University, Thanjavur

FEATURE EXTRACTION. Dr. K.Vijayarekha. Associate Dean School of Electrical and Electronics Engineering SASTRA University, Thanjavur FEATURE EXTRACTION Dr. K.Vjayarekha Assocate Dean School of Electrcal and Electroncs Engneerng SASTRA Unversty, Thanjavur613 41 Jont Intatve of IITs and IISc Funded by MHRD Page 1 of 8 Table of Contents

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

Feature Reduction and Selection

Feature Reduction and Selection Feature Reducton and Selecton Dr. Shuang LIANG School of Software Engneerng TongJ Unversty Fall, 2012 Today s Topcs Introducton Problems of Dmensonalty Feature Reducton Statstc methods Prncpal Components

More information

Online Detection and Classification of Moving Objects Using Progressively Improving Detectors

Online Detection and Classification of Moving Objects Using Progressively Improving Detectors Onlne Detecton and Classfcaton of Movng Objects Usng Progressvely Improvng Detectors Omar Javed Saad Al Mubarak Shah Computer Vson Lab School of Computer Scence Unversty of Central Florda Orlando, FL 32816

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

RECOGNIZING GENDER THROUGH FACIAL IMAGE USING SUPPORT VECTOR MACHINE

RECOGNIZING GENDER THROUGH FACIAL IMAGE USING SUPPORT VECTOR MACHINE Journal of Theoretcal and Appled Informaton Technology 30 th June 06. Vol.88. No.3 005-06 JATIT & LLS. All rghts reserved. ISSN: 99-8645 www.jatt.org E-ISSN: 87-395 RECOGNIZING GENDER THROUGH FACIAL IMAGE

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

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

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

Assignment # 2. Farrukh Jabeen Algorithms 510 Assignment #2 Due Date: June 15, 2009.

Assignment # 2. Farrukh Jabeen Algorithms 510 Assignment #2 Due Date: June 15, 2009. Farrukh Jabeen Algorthms 51 Assgnment #2 Due Date: June 15, 29. Assgnment # 2 Chapter 3 Dscrete Fourer Transforms Implement the FFT for the DFT. Descrbed n sectons 3.1 and 3.2. Delverables: 1. Concse descrpton

More information

Recognition of Handwritten Numerals Using a Combined Classifier with Hybrid Features

Recognition of Handwritten Numerals Using a Combined Classifier with Hybrid Features Recognton of Handwrtten Numerals Usng a Combned Classfer wth Hybrd Features Kyoung Mn Km 1,4, Joong Jo Park 2, Young G Song 3, In Cheol Km 1, and Chng Y. Suen 1 1 Centre for Pattern Recognton and Machne

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

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

12/2/2009. Announcements. Parametric / Non-parametric. Case-Based Reasoning. Nearest-Neighbor on Images. Nearest-Neighbor Classification

12/2/2009. Announcements. Parametric / Non-parametric. Case-Based Reasoning. Nearest-Neighbor on Images. Nearest-Neighbor Classification Introducton to Artfcal Intellgence V22.0472-001 Fall 2009 Lecture 24: Nearest-Neghbors & Support Vector Machnes Rob Fergus Dept of Computer Scence, Courant Insttute, NYU Sldes from Danel Yeung, John DeNero

More information

A Background Subtraction for a Vision-based User Interface *

A Background Subtraction for a Vision-based User Interface * A Background Subtracton for a Vson-based User Interface * Dongpyo Hong and Woontack Woo KJIST U-VR Lab. {dhon wwoo}@kjst.ac.kr Abstract In ths paper, we propose a robust and effcent background subtracton

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

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

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

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

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

Hermite Splines in Lie Groups as Products of Geodesics

Hermite Splines in Lie Groups as Products of Geodesics Hermte Splnes n Le Groups as Products of Geodescs Ethan Eade Updated May 28, 2017 1 Introducton 1.1 Goal Ths document defnes a curve n the Le group G parametrzed by tme and by structural parameters n the

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

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

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

SLAM Summer School 2006 Practical 2: SLAM using Monocular Vision

SLAM Summer School 2006 Practical 2: SLAM using Monocular Vision SLAM Summer School 2006 Practcal 2: SLAM usng Monocular Vson Javer Cvera, Unversty of Zaragoza Andrew J. Davson, Imperal College London J.M.M Montel, Unversty of Zaragoza. josemar@unzar.es, jcvera@unzar.es,

More information

Face Recognition University at Buffalo CSE666 Lecture Slides Resources:

Face Recognition University at Buffalo CSE666 Lecture Slides Resources: Face Recognton Unversty at Buffalo CSE666 Lecture Sldes Resources: http://www.face-rec.org/algorthms/ Overvew of face recognton algorthms Correlaton - Pxel based correspondence between two face mages Structural

More information

Available online at Available online at Advanced in Control Engineering and Information Science

Available online at   Available online at   Advanced in Control Engineering and Information Science Avalable onlne at wwwscencedrectcom Avalable onlne at wwwscencedrectcom Proceda Proceda Engneerng Engneerng 00 (2011) 15000 000 (2011) 1642 1646 Proceda Engneerng wwwelsevercom/locate/proceda Advanced

More information

Shape Representation Robust to the Sketching Order Using Distance Map and Direction Histogram

Shape Representation Robust to the Sketching Order Using Distance Map and Direction Histogram Shape Representaton Robust to the Sketchng Order Usng Dstance Map and Drecton Hstogram Department of Computer Scence Yonse Unversty Kwon Yun CONTENTS Revew Topc Proposed Method System Overvew Sketch Normalzaton

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

Analysis of Continuous Beams in General

Analysis of Continuous Beams in General Analyss of Contnuous Beams n General Contnuous beams consdered here are prsmatc, rgdly connected to each beam segment and supported at varous ponts along the beam. onts are selected at ponts of support,

More information

Description of NTU Approach to NTCIR3 Multilingual Information Retrieval

Description of NTU Approach to NTCIR3 Multilingual Information Retrieval Proceedngs of the Thrd NTCIR Workshop Descrpton of NTU Approach to NTCIR3 Multlngual Informaton Retreval Wen-Cheng Ln and Hsn-Hs Chen Department of Computer Scence and Informaton Engneerng Natonal Tawan

More information

TN348: Openlab Module - Colocalization

TN348: Openlab Module - Colocalization TN348: Openlab Module - Colocalzaton Topc The Colocalzaton module provdes the faclty to vsualze and quantfy colocalzaton between pars of mages. The Colocalzaton wndow contans a prevew of the two mages

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

Human Face Recognition Using Generalized. Kernel Fisher Discriminant

Human Face Recognition Using Generalized. Kernel Fisher Discriminant Human Face Recognton Usng Generalzed Kernel Fsher Dscrmnant ng-yu Sun,2 De-Shuang Huang Ln Guo. Insttute of Intellgent Machnes, Chnese Academy of Scences, P.O.ox 30, Hefe, Anhu, Chna. 2. Department of

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

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

Local Quaternary Patterns and Feature Local Quaternary Patterns

Local Quaternary Patterns and Feature Local Quaternary Patterns Local Quaternary Patterns and Feature Local Quaternary Patterns Jayu Gu and Chengjun Lu The Department of Computer Scence, New Jersey Insttute of Technology, Newark, NJ 0102, USA Abstract - Ths paper presents

More information

A Modified Median Filter for the Removal of Impulse Noise Based on the Support Vector Machines

A Modified Median Filter for the Removal of Impulse Noise Based on the Support Vector Machines A Modfed Medan Flter for the Removal of Impulse Nose Based on the Support Vector Machnes H. GOMEZ-MORENO, S. MALDONADO-BASCON, F. LOPEZ-FERRERAS, M. UTRILLA- MANSO AND P. GIL-JIMENEZ Departamento de Teoría

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

Image Representation & Visualization Basic Imaging Algorithms Shape Representation and Analysis. outline

Image Representation & Visualization Basic Imaging Algorithms Shape Representation and Analysis. outline mage Vsualzaton mage Vsualzaton mage Representaton & Vsualzaton Basc magng Algorthms Shape Representaton and Analyss outlne mage Representaton & Vsualzaton Basc magng Algorthms Shape Representaton and

More information

CSCI 5417 Information Retrieval Systems Jim Martin!

CSCI 5417 Information Retrieval Systems Jim Martin! CSCI 5417 Informaton Retreval Systems Jm Martn! Lecture 11 9/29/2011 Today 9/29 Classfcaton Naïve Bayes classfcaton Ungram LM 1 Where we are... Bascs of ad hoc retreval Indexng Term weghtng/scorng Cosne

More information

EXTENDED BIC CRITERION FOR MODEL SELECTION

EXTENDED BIC CRITERION FOR MODEL SELECTION IDIAP RESEARCH REPORT EXTEDED BIC CRITERIO FOR ODEL SELECTIO Itshak Lapdot Andrew orrs IDIAP-RR-0-4 Dalle olle Insttute for Perceptual Artfcal Intellgence P.O.Box 59 artgny Valas Swtzerland phone +4 7

More information

PRÉSENTATIONS DE PROJETS

PRÉSENTATIONS DE PROJETS PRÉSENTATIONS DE PROJETS Rex Onlne (V. Atanasu) What s Rex? Rex s an onlne browser for collectons of wrtten documents [1]. Asde ths core functon t has however many other applcatons that make t nterestng

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

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

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

Optimizing Document Scoring for Query Retrieval

Optimizing Document Scoring for Query Retrieval Optmzng Document Scorng for Query Retreval Brent Ellwen baellwe@cs.stanford.edu Abstract The goal of ths project was to automate the process of tunng a document query engne. Specfcally, I used machne learnng

More information

Online Handwritten Cursive Word Recognition Using Segmentation-free and Segmentation-based Methods

Online Handwritten Cursive Word Recognition Using Segmentation-free and Segmentation-based Methods 3th Asan Conference on Pattern Recognton (ACPR2015), November 3-6, 2015, Kuala Lumpur, Malasa. Onlne Handwrtten Cursve Word Recognton Usng Segmentaton-free and Segmentaton-based Methods Blan Zhua, Art

More information

CS246: Mining Massive Datasets Jure Leskovec, Stanford University

CS246: Mining Massive Datasets Jure Leskovec, Stanford University CS46: Mnng Massve Datasets Jure Leskovec, Stanford Unversty http://cs46.stanford.edu /19/013 Jure Leskovec, Stanford CS46: Mnng Massve Datasets, http://cs46.stanford.edu Perceptron: y = sgn( x Ho to fnd

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

Object-Based Techniques for Image Retrieval

Object-Based Techniques for Image Retrieval 54 Zhang, Gao, & Luo Chapter VII Object-Based Technques for Image Retreval Y. J. Zhang, Tsnghua Unversty, Chna Y. Y. Gao, Tsnghua Unversty, Chna Y. Luo, Tsnghua Unversty, Chna ABSTRACT To overcome the

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

Deep Classification in Large-scale Text Hierarchies

Deep Classification in Large-scale Text Hierarchies Deep Classfcaton n Large-scale Text Herarches Gu-Rong Xue Dkan Xng Qang Yang 2 Yong Yu Dept. of Computer Scence and Engneerng Shangha Jao-Tong Unversty {grxue, dkxng, yyu}@apex.sjtu.edu.cn 2 Hong Kong

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

CS 534: Computer Vision Model Fitting

CS 534: Computer Vision Model Fitting CS 534: Computer Vson Model Fttng Sprng 004 Ahmed Elgammal Dept of Computer Scence CS 534 Model Fttng - 1 Outlnes Model fttng s mportant Least-squares fttng Maxmum lkelhood estmaton MAP estmaton Robust

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

Machine Learning 9. week

Machine Learning 9. week Machne Learnng 9. week Mappng Concept Radal Bass Functons (RBF) RBF Networks 1 Mappng It s probably the best scenaro for the classfcaton of two dataset s to separate them lnearly. As you see n the below

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

Fast Feature Value Searching for Face Detection

Fast Feature Value Searching for Face Detection Vol., No. 2 Computer and Informaton Scence Fast Feature Value Searchng for Face Detecton Yunyang Yan Department of Computer Engneerng Huayn Insttute of Technology Hua an 22300, Chna E-mal: areyyyke@63.com

More information

UB at GeoCLEF Department of Geography Abstract

UB at GeoCLEF Department of Geography   Abstract UB at GeoCLEF 2006 Mguel E. Ruz (1), Stuart Shapro (2), June Abbas (1), Slva B. Southwck (1) and Davd Mark (3) State Unversty of New York at Buffalo (1) Department of Lbrary and Informaton Studes (2) Department

More information

NAG Fortran Library Chapter Introduction. G10 Smoothing in Statistics

NAG Fortran Library Chapter Introduction. G10 Smoothing in Statistics Introducton G10 NAG Fortran Lbrary Chapter Introducton G10 Smoothng n Statstcs Contents 1 Scope of the Chapter... 2 2 Background to the Problems... 2 2.1 Smoothng Methods... 2 2.2 Smoothng Splnes and Regresson

More information

A New Approach For the Ranking of Fuzzy Sets With Different Heights

A New Approach For the Ranking of Fuzzy Sets With Different Heights New pproach For the ankng of Fuzzy Sets Wth Dfferent Heghts Pushpnder Sngh School of Mathematcs Computer pplcatons Thapar Unversty, Patala-7 00 Inda pushpndersnl@gmalcom STCT ankng of fuzzy sets plays

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

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

USING GRAPHING SKILLS

USING GRAPHING SKILLS Name: BOLOGY: Date: _ Class: USNG GRAPHNG SKLLS NTRODUCTON: Recorded data can be plotted on a graph. A graph s a pctoral representaton of nformaton recorded n a data table. t s used to show a relatonshp

More information

Using Fuzzy Logic to Enhance the Large Size Remote Sensing Images

Using Fuzzy Logic to Enhance the Large Size Remote Sensing Images Internatonal Journal of Informaton and Electroncs Engneerng Vol. 5 No. 6 November 015 Usng Fuzzy Logc to Enhance the Large Sze Remote Sensng Images Trung Nguyen Tu Huy Ngo Hoang and Thoa Vu Van Abstract

More information

Fuzzy Modeling of the Complexity vs. Accuracy Trade-off in a Sequential Two-Stage Multi-Classifier System

Fuzzy Modeling of the Complexity vs. Accuracy Trade-off in a Sequential Two-Stage Multi-Classifier System Fuzzy Modelng of the Complexty vs. Accuracy Trade-off n a Sequental Two-Stage Mult-Classfer System MARK LAST 1 Department of Informaton Systems Engneerng Ben-Guron Unversty of the Negev Beer-Sheva 84105

More information

Signature and Lexicon Pruning Techniques

Signature and Lexicon Pruning Techniques Sgnature and Lexcon Prunng Technques Srnvas Palla, Hansheng Le, Venu Govndaraju Centre for Unfed Bometrcs and Sensors Unversty at Buffalo {spalla2, hle, govnd}@cedar.buffalo.edu Abstract Handwrtten word

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

SHAPE RECOGNITION METHOD BASED ON THE k-nearest NEIGHBOR RULE

SHAPE RECOGNITION METHOD BASED ON THE k-nearest NEIGHBOR RULE SHAPE RECOGNITION METHOD BASED ON THE k-nearest NEIGHBOR RULE Dorna Purcaru Faculty of Automaton, Computers and Electroncs Unersty of Craoa 13 Al. I. Cuza Street, Craoa RO-1100 ROMANIA E-mal: dpurcaru@electroncs.uc.ro

More information

Histogram of Template for Pedestrian Detection

Histogram of Template for Pedestrian Detection PAPER IEICE TRANS. FUNDAMENTALS/COMMUN./ELECTRON./INF. & SYST., VOL. E85-A/B/C/D, No. xx JANUARY 20xx Hstogram of Template for Pedestran Detecton Shaopeng Tang, Non Member, Satosh Goto Fellow Summary In

More information

Fuzzy C-Means Initialized by Fixed Threshold Clustering for Improving Image Retrieval

Fuzzy C-Means Initialized by Fixed Threshold Clustering for Improving Image Retrieval Fuzzy -Means Intalzed by Fxed Threshold lusterng for Improvng Image Retreval NAWARA HANSIRI, SIRIPORN SUPRATID,HOM KIMPAN 3 Faculty of Informaton Technology Rangst Unversty Muang-Ake, Paholyotn Road, Patumtan,

More information

Classification of Face Images Based on Gender using Dimensionality Reduction Techniques and SVM

Classification of Face Images Based on Gender using Dimensionality Reduction Techniques and SVM Classfcaton of Face Images Based on Gender usng Dmensonalty Reducton Technques and SVM Fahm Mannan 260 266 294 School of Computer Scence McGll Unversty Abstract Ths report presents gender classfcaton based

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

Segmentation Based Online Word Recognition: A Conditional Random Field Driven Beam Search Strategy

Segmentation Based Online Word Recognition: A Conditional Random Field Driven Beam Search Strategy 2013 12th Internatonal Conference on Document Analyss and Recognton Segmentaton Based Onlne Word Recognton: A Condtonal Random Feld Drven Beam Search Strategy Art Shvram 1, Blan Zhu 2, Srrangaraj Setlur

More information

Support Vector Machines

Support Vector Machines Support Vector Machnes Decson surface s a hyperplane (lne n 2D) n feature space (smlar to the Perceptron) Arguably, the most mportant recent dscovery n machne learnng In a nutshell: map the data to a predetermned

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

Learning Non-Linearly Separable Boolean Functions With Linear Threshold Unit Trees and Madaline-Style Networks

Learning Non-Linearly Separable Boolean Functions With Linear Threshold Unit Trees and Madaline-Style Networks In AAAI-93: Proceedngs of the 11th Natonal Conference on Artfcal Intellgence, 33-1. Menlo Park, CA: AAAI Press. Learnng Non-Lnearly Separable Boolean Functons Wth Lnear Threshold Unt Trees and Madalne-Style

More information

Sixth Indian Conference on Computer Vision, Graphics & Image Processing

Sixth Indian Conference on Computer Vision, Graphics & Image Processing Sxth Indan Conference on Computer Vson, Graphcs & Image Processng Incorporatng Cohort Informaton for Relable Palmprnt Authentcaton Ajay Kumar Bometrcs Research Laboratory, Department of Electrcal Engneerng

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

A Clustering Algorithm for Key Frame Extraction Based on Density Peak

A Clustering Algorithm for Key Frame Extraction Based on Density Peak Journal of Computer and Communcatons, 2018, 6, 118-128 http://www.scrp.org/ournal/cc ISSN Onlne: 2327-5227 ISSN Prnt: 2327-5219 A Clusterng Algorthm for Key Frame Extracton Based on Densty Peak Hong Zhao

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