A Chunking Strategy Towards Unknown Word Detection in Chinese Word Segmentation
|
|
- Louisa Harris
- 5 years ago
- Views:
Transcription
1 A Chukg Strategy Towards Ukow Word Detecto Chese Word Segmetato Zhou GuoDog Isttute for Ifocomm Research, 2 Heg Mu Keg Terrace, Sgapore 963 zhougd@2r.a-star.edu.sg Abstract. Ths paper proposes a chukg strategy to detect ukow words Chese word segmetato. Frst, a raw setece s pre-segmeted to a sequece of word atoms usg a maxmum matchg algorthm. The a chukg model s appled to detect ukow words by chukg oe or more word atoms together accordg to the word formato patters of the word atoms. I ths paper, a dscrmatve Markov model, amed Mutual Iformato Idepedece Model (MIIM), s adopted chukg. Besdes, a maxmum etropy model s appled to tegrate varous types of cotexts ad resolve the data sparseess problem MIIM. Moreover, a error-drve learg approach s proposed to lear useful cotexts the maxmum etropy model. I ths way, the umber of cotexts the maxmum etropy model ca be sgfcatly reduced wthout performace decrease. Ths makes t possble for further mprovg the performace by cosderg more varous types of cotexts. Evaluato o the PK ad CTB corpora the Frst SIGHAN Chese word segmetato bakeoff shows that our chukg approach successfully detects about 80% of ukow words o both of the corpora ad outperforms the best-reported systems by 8.% ad 7.% ukow word detecto o them respectvely. Itroducto Pror to ay lgustc aalyss of Chese text, Chese word segmetato s the ecessary frst step ad oe of major bottleecks Chese formato processg sce a Chese setece s wrtte a cotuous strg of characters wthout obvous separators (such as blaks) betwee the words. Durg the past two decades, ths research has bee a hot topc Chese formato processg [-0]. There exst two major problems Chese word segmetato: ambguty resoluto ad ukow word detecto. Whle -gram modelg ad/or word cooccurrece has bee successfully appled to deal wth the ambguty problems [3, 5, 0, 2, 3], ukow word detecto has become the major bottleeck Chese I ths paper, word atoms refer to basc buldg uts words. For example, the word Ï (computer) cossts of two word atoms: Ï (computg) ad (mache). Geerally, word atoms ca ether occur depedetly, e.g. Ï (computg), or oly become a part of a word, e.g. (mache) the word Ï (computer). R. Dale et al. (Eds.): IJCNLP 2005, LNAI 365, pp , Sprger-Verlag Berl Hedelberg 2005
2 A Chukg Strategy Towards Ukow Word Detecto 53 word segmetato. Curretly, almost all Chese word segmetato systems rely o a word dctoary. The problem s that whe the words stored the dctoary are suffcet, the system's performace wll be greatly deterorated by the presece of words that are ukow to the system. Moreover, maual mateace of a dctoary s very tedous ad tme cosumg. It s therefore mportat for a Chese word segmetato system to detfy ukow words from the text automatcally. I lterature, two categores of competg approaches are wdely used to detect ukow words 2 : statstcal approaches [5,, 2, 3, 4, 5] ad rule-based approaches [5,, 4, 5]. Although rule-based approaches have the advatage of beg smple, the complexty ad doma depedecy of how the ukow words are produced greatly reduce the effcecy of these approaches. O the other had, statstcal approaches have the advatage of beg doma-depedet [6]. It s terestg to ote that may systems apply a hybrd approach [5,, 4, 5]. Regardless of the choce of dfferet approaches, fdg a way to automatcally detect ukow words has become a crucal ssue Chese word segmetato ad Chese formato processg geeral. Iput raw setece: d ~. MMA pre-segmetato: À Ĺ ~. Ukow word detecto: À Ĺ ~. Zhag Je graduate from JaoTog Uversty. Fg.. MMA ad ukow word detecto by chukg: a example Ths paper proposes a chukg strategy to cope wth ukow words Chese word segmetato. Frst, a raw setece s pre-segmeted to a sequece of word atoms (.e. sgle-character words ad mult-character words) usg a maxmum matchg algorthm (MMA) 3. The a chukg model s appled to detect ukow words by chukg oe or more word atoms together accordg to the word formato patters of the word atoms. Fgure gves a example. Here, the problem of ukow word detecto s re-cast as chukg oe or more word atoms together to form a ew word ad a dscrmatve Markov model, amed Mutual Iformato Idepedece Model (MIIM), s adopted chukg. Besdes, a maxmum etropy model s appled to tegrate varous types of cotexts ad resolve the data sparseess problem MIIM. Moreover, a error-drve learg approach s proposed to lear useful 2 Some systems [3,4] focus o proper ames due to ther mportace Chese formato processg. 3 A typcal MMA detfes all character sequeces whch are foud the word dctoary ad marks them as words. Those character sequeces, whch ca be segmeted more tha oe way, are marked as ambguous ad a word ugram model s appled to choose the most lkely segmetato sequece. The remag sequeces,.e. those ot foud the dctoary, are called fragmets ad segmeted to sgle characters. I ths way, each Chese setece s pre-segmeted to a sequece of sgle-character words ad multcharacter words. For coveece, we call these sgle-character words ad mult-character words the output of the MMA algorthm as word atoms.
3 532 G. Zhou cotexts the maxmum etropy model. I ths way, the umber of cotexts the maxmum etropy model ca be sgfcatly reduced wthout performace decrease. Ths makes t possble for further mprovg the performace by cosderg more varous types of cotexts the future. Evaluato o the PK ad CTB corpora the Frst SIGHAN Chese word segmetato bakeoff shows that our chukg strategy performs best ukow word detecto o both of the corpora. The rest of the paper s as follows: I Secto 2, we wll dscuss detals about our chukg strategy ukow word detecto. Expermetal results are gve Secto 3. Fally, some remarks ad coclusos are made Secto 4. 2 Ukow Word Detecto by Chukg I ths secto, we wll frst descrbe the chukg strategy ukow word detecto of Chese word segmetato usg a dscrmatve Markov model, called Mutual Iformato Idepedece Model (MIIM). The a maxmum etropy model s appled to tegrate varous types of cotexts ad resolve the data sparseess problem MIIM. Fally, a error-drve learg approach s proposed to select useful cotexts ad reduce the cotext feature vector dmeso. 2. Mutual Iformato Idepedece Model ad Ukow Word Detecto Mutual Iformato Idepedece Model I ths paper, we use a dscrmatve Markov model, called Mutual Iformato Idepedece Model (MIIM) proposed by Zhou et al [7] 4, ukow word detecto by chukg. MIIM s derved from a codtoal probablty model. Gve a observato sequece O = o o 2 Lo, the goal of a codtoal probablty model s to fd a stochastc optmal state(tag) sequece S = s s 2 Ls that maxmzes: log P( S O P( S, O ) ) = log P( S ) + log () P( S ) P( O ) The secod term Equato () s the par-wse mutual formato (PMI) betwee S ad O. I order to smplfy the computato of ths term, we assume a par-wse mutual formato depedece (2):, O ) = PMI ( s, O ) = PMI( S or P( S, O ) P( s, O ) log = (2) P( S ) P( O ) ) log = P( s ) P( O 4 We have reamed the dscrmatve Markov model [7] as the Mutual Iformato Idepedece Model accordg to the ovel par-wse mutual formato depedece assumpto the model. Aother reaso s to dstgush t from the tradtoal Hdde Markov Model [8] ad avod msleadg.
4 A Chukg Strategy Towards Ukow Word Detecto 533 That s, a dvdual state s oly depedet o the observato sequece depedet o other states the state sequece O ad S. Ths assumpto s reasoable because the depedece amog the states the state sequece S has already bee captured by the frst term Equato (). Applyg Equato (2) to Equato (), we have Equato (3) 5 : O ) = PMI( s, S ) + log P( s O ) = 2 = log P( S (3) We call the above model as show Equato (3) the Mutual Iformato Idepedece Model due to ts par-wse mutual formato assumpto as show Equato (2). The above model cossts of two sub-models: the state trasto model = 2 = PMI ( s, S as the frst term Equato (3) ad the output model ) log P ( s O as the secod term Equato (3). Here, a varat of the Vterb ) algorthm [9] decodg the stadard Hdde Markov Model (HMM) [8] s mplemeted to fd the most lkely state sequece by replacg the state trasto model ad the output model of the stadard HMM wth the state trasto model ad the output model of the MIIM, respectvely. Ukow Word Detecto For ukow word detecto by chukg, a word (kow word or ukow word) s regarded as a chuk of oe or more word atoms ad we have: o =< p, w > ; w s the th word atom the sequece of word atomsw = w w 2 Lw ; p s the word formato patter of the word atom w. Here p measures the word formato power of the word atom w ad cossts of: o The percetage of w occurrg as a whole word (roud to 0%) o The percetage of w occurrg at the begg of other words (roud to 0%) o The percetage of w w occurrg at the ed of other words (roud to 0%) o The legth of o The occurrg frequecy feature of w, whch s mapped to max(log(frequecy), 9 ). s : the states are used to bracket ad dfferetate varous types of words. I ths way, Chese ukow word detecto ca be regarded as a bracketg process whle dfferetato of dfferet word types ca help the bracketg process. s s structural ad cossts of three parts: 5 Detals about the dervato are omtted due to space lmtato. Please see [7] for more.
5 534 G. Zhou o Boudary Category (B): t cludes four values: {O, B, M, E}, where O meas that curret word atom s a whole word ad B/M/E meas that curret word atom s at the Begg/ the Mddle/at the Ed of a word. o Word Category (W): It s used to deote the class of the word. I our system, words are classfed to two types: pure Chese word type ad mxed word type (.e. cludg Eglsh characters ad Chese dgts/umbers/symbols). o Word Atom Formato Patter (P): Because of the lmted umber of boudary ad word categores, the word atom formato patter descrbed above s added to the structural state to represet a more accurate state trasto model MIIM whle keepg ts output model. Problem wth Ukow Word Detecto Usg MIIM From Equato (3), we ca see that the state trasto model of MIIM ca be computed by usg gram modelg [20, 2, 22], where each tag s assumed to be depedet o the N- prevous tags (e.g. 2). The problem wth the above MIIM les the data sparseess problem rased by ts output model: log P ( s O = ). Ideally, we would have suffcet trag data for every evet whose codtoal probablty we wsh to calculate. Ufortuately, there s rarely eough trag data to compute accurate probabltes whe decodg o ew data. Geerally, two smoothg approaches [2, 22, 23] are appled to resolve ths problem: lear terpolato ad back-off. However, these two approaches oly work well whe the umber of dfferet formato sources s very lmted. Whe a few features ad/or a log cotext are cosdered, the umber of dfferet formato sources s expoetal. Ths makes smoothg approaches approprate our system. I ths paper, the maxmum etropy model [24] s proposed to tegrate varous cotext formato sources ad resolve the data sparseess problem our system. The reaso that we choose the maxmum etropy model for ths purpose s that t represets the state-of the-art the mache learg research commuty ad there are good mplemetatos of the algorthm avalable. Here, we use the ope NLP maxmum etropy package 6 our system. 2.2 Maxmum Etropy The maxmum etropy model s a probablty dstrbuto estmato techque wdely used recet years for atural laguage processg tasks. The prcple of the maxmum etropy model estmatg probabltes s to clude as much formato as s kow from the data whle makg o addtoal assumptos. The maxmum etropy model returs the probablty dstrbuto that satsfes the above property wth the hghest etropy. Formally, the decso fucto of the maxmum etropy model ca be represeted as: k f j ( h, o) P( o, h) = α (4) j Z ( h) j= 6
6 A Chukg Strategy Towards Ukow Word Detecto 535 where o s the outcome, h s the hstory (cotext feature vector ths paper), Z(h) s a ormalzato fucto, {f, f 2,..., f k } are feature fuctos ad {α, α 2,, α k } are the model parameters. Each model parameter correspods to exactly oe feature ad ca be vewed as a "weght" for that feature. All features used the maxmum etropy model are bary, e.g. f j, f o = Idepede tword, CurretWor datom = ( h, o) = 0, otherwse. y ( we ); I order to relably estmate P ( s O ) the output model of MIIM usg the maxmum etropy model, varous cotext formato sources are cluded the cotext feature vector: p : curret word atom formato patter : prevous word atom formato patter ad curret word atom formato patter p : curret word atom formato patter ad ext word atom formato patter p : curret word atom formato patter ad curret word atom p p p + w p w p : prevous word atom formato patter, prevous word atom ad curret word atom formato patter p p+ w+ : curret word atom formato patter, ext word atom formato patter ad ext word atom p pw : prevous word atom formato patter, curret word atom formato patter ad curret word atom p w p+ : curret word atom formato patter, curret word atom ad ext word atom formato patter p w p w : prevous word atom formato patter, prevous word atom, curret word atom formato patter ad curret word atom p w p+ w+ : curret word atom formato patter, curret word atom, ext word atom formato patter ad ext word atom However, there exsts a problem whe we clude above varous cotext formato the maxmum etropy model: the cotext feature vector dmeso easly becomes too large for the model to hadle. Oe easy soluto to ths problem s to oly keep those frequetly occurrg cotexts the model. Although ths frequecy flterg approach s smple, may useful cotexts may ot occur frequetly ad be fltered out whle those kept may ot be useful. To resolve ths problem, we propose a alteratve error-drve learg approach to oly keep useful cotexts the model. 2.3 Cotext Feature Selecto Usg Error-Drve Learg Here, we propose a error-drve learg approach to exame the effectveess of varous cotexts ad select useful cotexts to reduce the sze of the cotext feature (5)
7 536 G. Zhou vector used the maxmum etropy model for estmatg P ( s O ) the output model of MIIM. Ths makes t possble to further mprove the performace by corporatg more varous types of cotexts the future. Assume Φ s the cotaer for useful cotexts. Gve a set of exstg useful cotexts Φ ad a set of ew cotexts Φ, the effectveess of a ew cotext C Φ, E( Φ, C ), s measured by the C -related reducto errors whch results from addg the ew cotext set Φ to the useful cotext set Φ : E Φ, C ) = # Error( Φ, C ) # Error( Φ + Φ, C ) (6) ( Here, # Error( Φ, C ) s the umber of C -related chukg errors before Φ s added to Φ ad # Error( Φ + Φ, C ) s the umber of C -related chukg errors after Φ s added to Φ. That s, E( Φ, C ) s the umber of the chukg error correctos made o the cotext C Φ whe Φ s added to Φ. If E ( Φ, C ) > 0, we declare that the ew cotext C s a useful cotext ad should be added to Φ. Otherwse, the ew cotext C s cosdered useless ad dscarded. Gve the above error-drve learg approach, we talze Φ = { p } (.e. we assume all the curret word atom formato patters are useful cotexts) ad choose oe of the other cotext types as the ew cotext set Φ, e.g. Φ = { p w }. The, we ca tra two MIIMs wth dfferet output models usg Φ ad Φ + Φ respectvely. Moreover, useful cotexts are leart o the trag data a two-fold way. For each fold, two MIIMs are traed o 50% of the trag data ad for each ew cotext C Φ, evaluate ts effectveess E( Φ, C ) o the remag 50% of the trag data accordg to the cotext effectveess measure as show Equato (6). If E ( Φ, C ) > 0, C s marked as a useful cotext ad added to Φ. I ths way, all the useful cotexts Φ are corporated to the useful cotext set Φ. Smlarly, we ca clude useful cotexts of other cotext types to the useful cotext set Φ oe by oe. I ths paper, varous types of cotexts are leart oe by oe the exact same order as show Secto 2.2. Fally, sce dfferet types of cotexts may have cross-effects, the above process s terated wth the reewed useful cotext set Φ utl very few useful cotexts ca be foud at each loop. Our expermets show that terato coverges wth four loops. 3 Expermetal Results All of our expermets are evaluated o the PK ad CTB bechmark corpora used the Frst SIGHAN Chese word segmetato bakeoff 7 wth the closed cofgurato. That s, oly the trag data from the partcular corpus s used durg trag. For ukow word detecto, the chukg trag data s derved by usg the same Maxmum Matchg Algorthm (MMA) to segmet each word the orgal trag data as a chuk of word atoms. Ths s doe a two-fold way. For each fold, the 7
8 A Chukg Strategy Towards Ukow Word Detecto 537 MMA s traed o 50% of the orgal trag data ad the used to segmet the remag 50% of the orgal trag data. The the MIIM s used to tra a chukg model for ukow word detecto o the chukg trag data. Table shows the detals of the two corpora. Here, s defed as the percetage of words the test corpus ot occurrg the trag corpus ad dcates the out-ofvocabulary rate the test corpus. Table. Statstcs of the corpora used our evaluato Corpus Abbrevato Trag Data Test Data Bejg Uversty PK 6.9% 00K words 7K words UPENN Chese Treebak CTB 8.% 250K words 40K words Table 2 shows the detaled performace of our system ukow word detecto ad Chese word segmetato as a whole usg the stadard scorg scrpt 8 o the test data. I ths ad subsequet tables, varous evaluato measures are provded: precso (P), recall (R), F-measure, recall o out-of-vocabulary words ( R ) ad recall o -vocabulary words ( R IV ). It shows that our system acheves precso/recall/f-measure of 93.5%/96.%/94.8 ad 90.5%/90.%/90.3 o the PK ad CTB corpora respectvely. Especally, our chukg approach ca successfully detect 80.5% ad 77.6% of ukow words o the PK ad CTB corpora respectvely. Table 2. Detaled performace of our system o the st SIGHAN Chese word segmetato bechmark data Corpus P R F R R IV PK CTB Table 3 ad Table 4 compare our system wth other best-reported systems o the PK ad CTB corpora respectvely. Table 3 shows that our chukg approach ukow word detecto outperforms others by more tha 8% o the PK corpus. It also shows that our system performs comparably wth the best reported systems o the PK corpus whe the out-of-vocabulary rate s moderate(6.9%). Our performace Chese word segmetato as a whole s somewhat pulled dow by the lower performace recallg -vocabulary words. Ths may be due to the preferece of our chukg strategy detectg ukow words by wrogly combg some of vocabulary words to ukow words. Such preferece may cause egatve effect Chese word segmetato as a whole whe the ga ukow word detecto fals to compesate the loss wrogly combg some of -vocabulary words to ukow words. Ths happes whe the out-of-vocabulary rate s ot hgh, e.g. o the 8
9 538 G. Zhou PK corpus. Table 4 shows that our chukg approach ukow word detecto outperforms others by more tha 7% o the CTB corpus. It also shows that our system outperforms the other best-reported systems by more tha 2% Chese word segmetato as a whole o the CTB corpus. Ths s largely due to the huge ga ukow word detecto whe the out-of-vocabulary rate s hgh (e.g. 8.% the CTB corpus), eve though our system performs worse o recallg -vocabulary words tha others. Evaluato o both the PK ad CTB corpora shows that our chukg approach ca successfully detect about 80% of ukow words o corpora wth a large rage of the out-of-vocabulary rates. Ths suggests the powerfuless of usg varous word formato patters of word atoms detectg ukow words. Ths also demostrates the effectveess ad robustess of our chukg approach ukow word detecto of Chese word segmetato ad ts portablty to dfferet geres. Table 3. Comparso of our system wth other best-reported systems o the PK corpus Corpus P R F R R IV Ours Zhag et al [25] Wu [26] Che [27] Table 4. Comparso of our system wth other best-reported systems o the CTB corpus Corpus P R F R R IV Ours Zhag et al [25] Dua et al [28] Fally, Table 5 ad Table 6 compare our error-drve learg approach wth the frequecy flterg approach learg useful cotexts for the output model of MIIM o the PK ad CTB corpora respectvely. Due to memory lmtato, at most 400K useful cotexts are cosdered the frequecy flterg approach. Frst, they show that the error-drve learg approach s much more effectve tha the smple frequecy flterg approach. Wth the same umber of useful cotexts, the errordrve learg approach outperforms the frequecy flterg approach by 7.8%/0.6% ad 5.5%/0.8% R (ukow word detecto)/f-measure(chese word segmetato as a whole) o the PK ad CTB corpora respectvely. Moreover, the error-drve learg approach slghtly outperforms the frequecy flterg approach wth the best cofgurato of 2.5 ad 3.5 tmes of useful cotexts. Secod, they show that creasg the umber of frequetly occurrg cotexts usg the frequecy flterg approach may ot crease the performace. Ths may be due to that some of frequetly occurrg cotexts are osy or useless ad cludg them may have
10 A Chukg Strategy Towards Ukow Word Detecto 539 egatve effect. Thrd, they show that the error-drve learg approach s effectve learg useful cotexts by reducg 96-98% of possble cotexts. Fally, the fgures sde paretheses show the umber of useful patters shared betwee the error-drve learg approach ad the frequecy flterg approach. They show that about 40-50% of useful cotexts selected usg the error-drve learg approach do ot occur frequetly the useful cotexts selected usg the frequecy flterg approach. Table 5. Comparso of the error-drve learg approach wth the frequecy flterg approach learg useful cotexts for the output model of MIIM o the PK corpus (Total umber of possble cotexts: 4836K) Approach #useful cotexts F R R IV Error-Drve Learg 98K Frequecy Flterg 98K (63K) Frequecy Flterg (best performace) 250K (90K) Frequecy Flterg 400K (94K) Table 6. Comparso of the error-drve learg approach wth the frequecy flterg approach learg useful cotexts for the output model of MIIM o the CTB corpus (Total umber of possble cotexts: 038K) Approach #useful cotexts F R R IV Error-Drve Learg 43K Frequecy Flterg 43K (2K) Frequecy Flterg (best performace) 50K Frequecy Flterg 400K (40K) Cocluso I ths paper, a chukg strategy s preseted to detect ukow words Chese word segmetato by chukg oe or more word atoms together accordg to the varous word formato patters of the word atoms. Besdes, a maxmum etropy model s appled to tegrate varous types of cotexts ad resolve the data sparseess problem our strategy. Fally, a error-drve learg approach s proposed to lear useful cotexts the maxmum etropy model. I ths way, the umber of cotexts the maxmum etropy model ca be sgfcatly reduced wthout performace decrease. Ths makes t possble for further mprovg the performace by cosderg more varous types of cotexts. Evaluato o the PK ad CTB corpora the Frst SIGHAN Chese word segmetato bakeoff shows that our chukg strategy ca detect about 80% of ukow words o both of the corpora ad outperforms the best-reported systems by 8.% ad 7.% ukow word detecto
11 540 G. Zhou o them respectvely. Whle our Chese word segmetato system wth chukgbased ukow word detecto performs comparably wth the best systems o the PK corpus whe the out-of-vocabulary rate s moderate(6.9%), our system sgfcatly outperforms others by more tha 2% whe the out-of-vocabulary rate s hgh(8.%). Ths demostrates the effectveess ad robustess of our chukg strategy ukow word detecto of Chese word segmetato ad ts portablty to dfferet geres. Refereces. Je CY, Lu Y ad Lag NY. (989). O methods of Chese automatc segmetato, Joural of Chese Iformato Processg, 3(): L KC, Lu KY ad Zhag YK. (988). Segmetg Chese word ad processg dfferet meags structure, Joural of Chese Iformato Processg, 2(3): Lag NY, (990). The kowledge of Chese word segmetato, Joural of Chese Iformato Processg, 4(2): Lua KT, (990). From character to word - A applcato of formato theory, Computer Processg of Chese & Oretal Laguages, 4(4): Lua KT ad Ga GW. (994). A applcato of formato theory Chese word segmetato. Computer Processg of Chese & Oretal Laguages, 8(): Wag YC, SU HJ ad Mo Y. (990). Automatc processg of Chese words. Joural of Chese Iformato Processg. 4(4):-. 7. Wu JM ad Tseg G. (993). Chese text segmetato for text retreval: achevemets ad problems. Joural of the Amerca Socety for Iformato Scece. 44(9): Xu H, He KK ad Su B. (99) The mplemetato of a wrtte Chese automatc segmetato expert system, Joural of Chese Iformato Processg, 5(3): Yao TS, Zhag GP ad Wu YM. (990). A rule-based Chese automatc segmetato system, Joural of Chese Iformato Processg, 4(): Yeh CL ad Lee HJ. (995). Rule-based word detfcato for Madar Chese seteces - A ufcato approach, Computer Processg of Chese & Oretal Laguages, 9(2): Ne JY, J WY ad Mare-Louse Haa. (997). A hybrd approach to ukow word detecto ad segmetato of Chese, Chese Processg of Chese ad Oretal Laguages, (4): pp Tug CH ad Lee HJ. (994). Idetfcato of ukow word from a corpus, computer Processg of Chese & Oretal Laguages, 8(Supplemet): Chag JS et al. (994). A mult-corpus approach to recogto of proper ames Chese Text, Computer Processg of Chese & Oretal Laguages, 8(): Su MS, Huag CN, Gao HY ad Fag J. (994). Idetfyg Chese Names I Urestrcted Texts, Commucatos of Chese ad Oretal Laguages Iformato Processg Socety, 4(2): Zhou GD ad Lua KT, (997). Detecto of Ukow Chese Words Usg a Hybrd Approach, Computer Processg of Chese & Oretal Laguage, (): Eugee Charak, Statstcal laguage learg, The MIT Press, ISBN Zhou GDog ad Su J. (2002). Named Etty Recogto Usg a HMM-based Chuk Tagger, Proceedgs of the Coferece o Aual Meetg for Computatoal Lgustcs (ACL 2002) , Phladelpha.
12 A Chukg Strategy Towards Ukow Word Detecto Raber L A Tutoral o Hdde Markov Models ad Selected Applcatos Speech Recogto. IEEE 77(2), pages Vterb A.J Error Bouds for Covolutoal Codes ad a Asymptotcally Optmum Decodg Algorthm. IEEE Trasactos o Iformato Theory, IT 3(2), Gale W.A. ad Sampso G Good-Turg frequecy estmato wthout tears. Joural of Quattatve Lgustcs. 2: Jelek F. (989). Self-Orgazed Laguage Modelg for Speech Recogto. I Alex Wabel ad Ka-Fu Lee(Edtors). Readgs Speech Recogtop. Morga Kaufma Katz S.M. (987). Estmato of Probabltes from Sparse Data for the Laguage Model Compoet of a Speech Recogzer. IEEE Trasactos o Acoustcs. Speech ad Sgal Processg. 35: Che ad Goodma. (996). A Emprcal Study of Smoothg Techques for Laguage Modelg. I Proceedgs of the 34th Aual Meetg of the Assocato of Computatoal Lgustcs (ACL 996). pp Sata Cruz, Calfora, USA. 24. Rataparkh A. (996). A Maxmum Etropy Model for Part-of-Speech Taggg. Proceedgs of the Coferece o Emprcal Methods Natural Laguage Processg., Zhag HP, Yu HK, Xog DY ad Lu Q. (2003). HHMM-based Chese Lexcal Aalyzer ICTCLAS. Proceedgs of 2 d SIGHAN Workshop o Chese Laguage Processg Sapporo, Japa. 26. Wu AD. (2003). Chese Word Segmetato MSR-NLP. Proceedgs of 2 d SIGHAN Workshop o Chese Laguage Processg Sapporo, Japa. 27. Che AT. (2003). Chese Word Segmetato Usg Mmal Lgustc Kowledge. Proceedgs of 2 d SIGHAN Workshop o Chese Laguage Processg Sapporo, Japa. 28. Dua HM, Ba XJ, Chag BB ad Yu SW. (2003). Chese Word Segmetato at Pekg Uversty. Proceedgs of 2 d SIGHAN Workshop o Chese Laguage Processg Sapporo, Japa.
Collocation Extraction Using Square Mutual Information Approaches. Received December 2010; revised January 2011
Iteratoal Joural of Kowledge www.jklp.org ad Laguage Processg KLP Iteratoal c2011 ISSN 2191-2734 Volume 2, Number 1, Jauary 2011 pp. 53-58 Collocato Extracto Usg Square Mutual Iformato Approaches Huaru
More informationBayes (Naïve or not) Classifiers: Generative Approach
Logstc regresso Bayes (Naïve or ot) Classfers: Geeratve Approach What do we mea by Geeratve approach: Lear p(y), p(x y) ad the apply bayes rule to compute p(y x) for makg predctos Ths s essetally makg
More informationIntroduction to local (nonparametric) density estimation. methods
Itroducto to local (oparametrc) desty estmato methods A slecture by Yu Lu for ECE 66 Sprg 014 1. Itroducto Ths slecture troduces two local desty estmato methods whch are Parze desty estmato ad k-earest
More informationA New Family of Transformations for Lifetime Data
Proceedgs of the World Cogress o Egeerg 4 Vol I, WCE 4, July - 4, 4, Lodo, U.K. A New Famly of Trasformatos for Lfetme Data Lakhaa Watthaacheewakul Abstract A famly of trasformatos s the oe of several
More informationApplication of Calibration Approach for Regression Coefficient Estimation under Two-stage Sampling Design
Authors: Pradp Basak, Kaustav Adtya, Hukum Chadra ad U.C. Sud Applcato of Calbrato Approach for Regresso Coeffcet Estmato uder Two-stage Samplg Desg Pradp Basak, Kaustav Adtya, Hukum Chadra ad U.C. Sud
More informationKernel-based Methods and Support Vector Machines
Kerel-based Methods ad Support Vector Maches Larr Holder CptS 570 Mache Learg School of Electrcal Egeerg ad Computer Scece Washgto State Uverst Refereces Muller et al. A Itroducto to Kerel-Based Learg
More informationUnsupervised Learning and Other Neural Networks
CSE 53 Soft Computg NOT PART OF THE FINAL Usupervsed Learg ad Other Neural Networs Itroducto Mture Destes ad Idetfablty ML Estmates Applcato to Normal Mtures Other Neural Networs Itroducto Prevously, all
More informationA tighter lower bound on the circuit size of the hardest Boolean functions
Electroc Colloquum o Computatoal Complexty, Report No. 86 2011) A tghter lower boud o the crcut sze of the hardest Boolea fuctos Masak Yamamoto Abstract I [IPL2005], Fradse ad Mlterse mproved bouds o the
More informationEstimation of Stress- Strength Reliability model using finite mixture of exponential distributions
Iteratoal Joural of Computatoal Egeerg Research Vol, 0 Issue, Estmato of Stress- Stregth Relablty model usg fte mxture of expoetal dstrbutos K.Sadhya, T.S.Umamaheswar Departmet of Mathematcs, Lal Bhadur
More informationMedian as a Weighted Arithmetic Mean of All Sample Observations
Meda as a Weghted Arthmetc Mea of All Sample Observatos SK Mshra Dept. of Ecoomcs NEHU, Shllog (Ida). Itroducto: Iumerably may textbooks Statstcs explctly meto that oe of the weakesses (or propertes) of
More informationBounds on the expected entropy and KL-divergence of sampled multinomial distributions. Brandon C. Roy
Bouds o the expected etropy ad KL-dvergece of sampled multomal dstrbutos Brado C. Roy bcroy@meda.mt.edu Orgal: May 18, 2011 Revsed: Jue 6, 2011 Abstract Iformato theoretc quattes calculated from a sampled
More informationBayes Estimator for Exponential Distribution with Extension of Jeffery Prior Information
Malaysa Joural of Mathematcal Sceces (): 97- (9) Bayes Estmator for Expoetal Dstrbuto wth Exteso of Jeffery Pror Iformato Hadeel Salm Al-Kutub ad Noor Akma Ibrahm Isttute for Mathematcal Research, Uverst
More informationSolving Constrained Flow-Shop Scheduling. Problems with Three Machines
It J Cotemp Math Sceces, Vol 5, 2010, o 19, 921-929 Solvg Costraed Flow-Shop Schedulg Problems wth Three Maches P Pada ad P Rajedra Departmet of Mathematcs, School of Advaced Sceces, VIT Uversty, Vellore-632
More informationFunctions of Random Variables
Fuctos of Radom Varables Chapter Fve Fuctos of Radom Varables 5. Itroducto A geeral egeerg aalyss model s show Fg. 5.. The model output (respose) cotas the performaces of a system or product, such as weght,
More informationPTAS for Bin-Packing
CS 663: Patter Matchg Algorthms Scrbe: Che Jag /9/00. Itroducto PTAS for B-Packg The B-Packg problem s NP-hard. If we use approxmato algorthms, the B-Packg problem could be solved polyomal tme. For example,
More informationNP!= P. By Liu Ran. Table of Contents. The P versus NP problem is a major unsolved problem in computer
NP!= P By Lu Ra Table of Cotets. Itroduce 2. Prelmary theorem 3. Proof 4. Expla 5. Cocluso. Itroduce The P versus NP problem s a major usolved problem computer scece. Iformally, t asks whether a computer
More informationLecture 02: Bounding tail distributions of a random variable
CSCI-B609: A Theorst s Toolkt, Fall 206 Aug 25 Lecture 02: Boudg tal dstrbutos of a radom varable Lecturer: Yua Zhou Scrbe: Yua Xe & Yua Zhou Let us cosder the ubased co flps aga. I.e. let the outcome
More informationComparing Different Estimators of three Parameters for Transmuted Weibull Distribution
Global Joural of Pure ad Appled Mathematcs. ISSN 0973-768 Volume 3, Number 9 (207), pp. 55-528 Research Ida Publcatos http://www.rpublcato.com Comparg Dfferet Estmators of three Parameters for Trasmuted
More informationNP!= P. By Liu Ran. Table of Contents. The P vs. NP problem is a major unsolved problem in computer
NP!= P By Lu Ra Table of Cotets. Itroduce 2. Strategy 3. Prelmary theorem 4. Proof 5. Expla 6. Cocluso. Itroduce The P vs. NP problem s a major usolved problem computer scece. Iformally, t asks whether
More information(b) By independence, the probability that the string 1011 is received correctly is
Soluto to Problem 1.31. (a) Let A be the evet that a 0 s trasmtted. Usg the total probablty theorem, the desred probablty s P(A)(1 ɛ ( 0)+ 1 P(A) ) (1 ɛ 1)=p(1 ɛ 0)+(1 p)(1 ɛ 1). (b) By depedece, the probablty
More informationPart 4b Asymptotic Results for MRR2 using PRESS. Recall that the PRESS statistic is a special type of cross validation procedure (see Allen (1971))
art 4b Asymptotc Results for MRR usg RESS Recall that the RESS statstc s a specal type of cross valdato procedure (see Alle (97)) partcular to the regresso problem ad volves fdg Y $,, the estmate at the
More informationBlock-Based Compact Thermal Modeling of Semiconductor Integrated Circuits
Block-Based Compact hermal Modelg of Semcoductor Itegrated Crcuts Master s hess Defese Caddate: Jg Ba Commttee Members: Dr. Mg-Cheg Cheg Dr. Daqg Hou Dr. Robert Schllg July 27, 2009 Outle Itroducto Backgroud
More informationLecture 9: Tolerant Testing
Lecture 9: Tolerat Testg Dael Kae Scrbe: Sakeerth Rao Aprl 4, 07 Abstract I ths lecture we prove a quas lear lower boud o the umber of samples eeded to do tolerat testg for L dstace. Tolerat Testg We have
More informationSummary of the lecture in Biostatistics
Summary of the lecture Bostatstcs Probablty Desty Fucto For a cotuos radom varable, a probablty desty fucto s a fucto such that: 0 dx a b) b a dx A probablty desty fucto provdes a smple descrpto of the
More informationA New Measure of Probabilistic Entropy. and its Properties
Appled Mathematcal Sceces, Vol. 4, 200, o. 28, 387-394 A New Measure of Probablstc Etropy ad ts Propertes Rajeesh Kumar Departmet of Mathematcs Kurukshetra Uversty Kurukshetra, Ida rajeesh_kuk@redffmal.com
More informationFeature Selection: Part 2. 1 Greedy Algorithms (continued from the last lecture)
CSE 546: Mache Learg Lecture 6 Feature Selecto: Part 2 Istructor: Sham Kakade Greedy Algorthms (cotued from the last lecture) There are varety of greedy algorthms ad umerous amg covetos for these algorthms.
More informationEconometric Methods. Review of Estimation
Ecoometrc Methods Revew of Estmato Estmatg the populato mea Radom samplg Pot ad terval estmators Lear estmators Ubased estmators Lear Ubased Estmators (LUEs) Effcecy (mmum varace) ad Best Lear Ubased Estmators
More informationTESTS BASED ON MAXIMUM LIKELIHOOD
ESE 5 Toy E. Smth. The Basc Example. TESTS BASED ON MAXIMUM LIKELIHOOD To llustrate the propertes of maxmum lkelhood estmates ad tests, we cosder the smplest possble case of estmatg the mea of the ormal
More informationA Method for Damping Estimation Based On Least Square Fit
Amerca Joural of Egeerg Research (AJER) 5 Amerca Joural of Egeerg Research (AJER) e-issn: 3-847 p-issn : 3-936 Volume-4, Issue-7, pp-5-9 www.ajer.org Research Paper Ope Access A Method for Dampg Estmato
More informationChapter 5 Properties of a Random Sample
Lecture 6 o BST 63: Statstcal Theory I Ku Zhag, /0/008 Revew for the prevous lecture Cocepts: t-dstrbuto, F-dstrbuto Theorems: Dstrbutos of sample mea ad sample varace, relatoshp betwee sample mea ad sample
More informationInterpolated Markov Models for Gene Finding
Iterpolated Markov Models for Gee Fdg BMI/CS 776 www.bostat.wsc.edu/bm776/ Sprg 2009 Mark Crave crave@bostat.wsc.edu The Gee Fdg Task Gve: a ucharacterzed DNA sequece Do: locate the gees the sequece, cludg
More informationChapter 4 (Part 1): Non-Parametric Classification (Sections ) Pattern Classification 4.3) Announcements
Aoucemets No-Parametrc Desty Estmato Techques HW assged Most of ths lecture was o the blacboard. These sldes cover the same materal as preseted DHS Bometrcs CSE 90-a Lecture 7 CSE90a Fall 06 CSE90a Fall
More informationAnalysis of Lagrange Interpolation Formula
P IJISET - Iteratoal Joural of Iovatve Scece, Egeerg & Techology, Vol. Issue, December 4. www.jset.com ISS 348 7968 Aalyss of Lagrage Iterpolato Formula Vjay Dahya PDepartmet of MathematcsMaharaja Surajmal
More informationLecture 3. Sampling, sampling distributions, and parameter estimation
Lecture 3 Samplg, samplg dstrbutos, ad parameter estmato Samplg Defto Populato s defed as the collecto of all the possble observatos of terest. The collecto of observatos we take from the populato s called
More information9.1 Introduction to the probit and logit models
EC3000 Ecoometrcs Lecture 9 Probt & Logt Aalss 9. Itroducto to the probt ad logt models 9. The logt model 9.3 The probt model Appedx 9. Itroducto to the probt ad logt models These models are used regressos
More informationPoint Estimation: definition of estimators
Pot Estmato: defto of estmators Pot estmator: ay fucto W (X,..., X ) of a data sample. The exercse of pot estmato s to use partcular fuctos of the data order to estmate certa ukow populato parameters.
More informationBayesian Classification. CS690L Data Mining: Classification(2) Bayesian Theorem: Basics. Bayesian Theorem. Training dataset. Naïve Bayes Classifier
Baa Classfcato CS6L Data Mg: Classfcato() Referece: J. Ha ad M. Kamber, Data Mg: Cocepts ad Techques robablstc learg: Calculate explct probabltes for hypothess, amog the most practcal approaches to certa
More informationAnalysis of Variance with Weibull Data
Aalyss of Varace wth Webull Data Lahaa Watthaacheewaul Abstract I statstcal data aalyss by aalyss of varace, the usual basc assumptos are that the model s addtve ad the errors are radomly, depedetly, ad
More information2.28 The Wall Street Journal is probably referring to the average number of cubes used per glass measured for some population that they have chosen.
.5 x 54.5 a. x 7. 786 7 b. The raked observatos are: 7.4, 7.5, 7.7, 7.8, 7.9, 8.0, 8.. Sce the sample sze 7 s odd, the meda s the (+)/ 4 th raked observato, or meda 7.8 c. The cosumer would more lkely
More informationCHAPTER VI Statistical Analysis of Experimental Data
Chapter VI Statstcal Aalyss of Expermetal Data CHAPTER VI Statstcal Aalyss of Expermetal Data Measuremets do ot lead to a uque value. Ths s a result of the multtude of errors (maly radom errors) that ca
More informationA Robust Total Least Mean Square Algorithm For Nonlinear Adaptive Filter
A Robust otal east Mea Square Algorthm For Nolear Adaptve Flter Ruxua We School of Electroc ad Iformato Egeerg X'a Jaotog Uversty X'a 70049, P.R. Cha rxwe@chare.com Chogzhao Ha, azhe u School of Electroc
More informationESTIMATION OF MISCLASSIFICATION ERROR USING BAYESIAN CLASSIFIERS
Producto Systems ad Iformato Egeerg Volume 5 (2009), pp. 4-50. ESTIMATION OF MISCLASSIFICATION ERROR USING BAYESIAN CLASSIFIERS PÉTER BARABÁS Uversty of Msolc, Hugary Departmet of Iformato Techology barabas@t.u-msolc.hu
More informationENGI 3423 Simple Linear Regression Page 12-01
ENGI 343 mple Lear Regresso Page - mple Lear Regresso ometmes a expermet s set up where the expermeter has cotrol over the values of oe or more varables X ad measures the resultg values of aother varable
More information1. BLAST (Karlin Altschul) Statistics
Parwse seuece algmet global ad local Multple seuece algmet Substtuto matrces Database searchg global local BLAST Seuece statstcs Evolutoary tree recostructo Gee Fdg Prote structure predcto RNA structure
More informationCIS 800/002 The Algorithmic Foundations of Data Privacy October 13, Lecture 9. Database Update Algorithms: Multiplicative Weights
CIS 800/002 The Algorthmc Foudatos of Data Prvacy October 13, 2011 Lecturer: Aaro Roth Lecture 9 Scrbe: Aaro Roth Database Update Algorthms: Multplcatve Weghts We ll recall aga) some deftos from last tme:
More informationAn Introduction to. Support Vector Machine
A Itroducto to Support Vector Mache Support Vector Mache (SVM) A classfer derved from statstcal learg theory by Vapk, et al. 99 SVM became famous whe, usg mages as put, t gave accuracy comparable to eural-etwork
More informationChapter 3 Sampling For Proportions and Percentages
Chapter 3 Samplg For Proportos ad Percetages I may stuatos, the characterstc uder study o whch the observatos are collected are qualtatve ature For example, the resposes of customers may marketg surveys
More informationA New Method for Decision Making Based on Soft Matrix Theory
Joural of Scetfc esearch & eports 3(5): 0-7, 04; rtcle o. JS.04.5.00 SCIENCEDOMIN teratoal www.scecedoma.org New Method for Decso Mag Based o Soft Matrx Theory Zhmg Zhag * College of Mathematcs ad Computer
More information2006 Jamie Trahan, Autar Kaw, Kevin Martin University of South Florida United States of America
SOLUTION OF SYSTEMS OF SIMULTANEOUS LINEAR EQUATIONS Gauss-Sedel Method 006 Jame Traha, Autar Kaw, Kev Mart Uversty of South Florda Uted States of Amerca kaw@eg.usf.edu Itroducto Ths worksheet demostrates
More informationComparison of Dual to Ratio-Cum-Product Estimators of Population Mean
Research Joural of Mathematcal ad Statstcal Sceces ISS 30 6047 Vol. 1(), 5-1, ovember (013) Res. J. Mathematcal ad Statstcal Sc. Comparso of Dual to Rato-Cum-Product Estmators of Populato Mea Abstract
More informationAnalyzing Fuzzy System Reliability Using Vague Set Theory
Iteratoal Joural of Appled Scece ad Egeerg 2003., : 82-88 Aalyzg Fuzzy System Relablty sg Vague Set Theory Shy-Mg Che Departmet of Computer Scece ad Iformato Egeerg, Natoal Tawa versty of Scece ad Techology,
More informationPrincipal Components. Analysis. Basic Intuition. A Method of Self Organized Learning
Prcpal Compoets Aalss A Method of Self Orgazed Learg Prcpal Compoets Aalss Stadard techque for data reducto statstcal patter matchg ad sgal processg Usupervsed learg: lear from examples wthout a teacher
More informationbest estimate (mean) for X uncertainty or error in the measurement (systematic, random or statistical) best
Error Aalyss Preamble Wheever a measuremet s made, the result followg from that measuremet s always subject to ucertaty The ucertaty ca be reduced by makg several measuremets of the same quatty or by mprovg
More informationAnalysis of System Performance IN2072 Chapter 5 Analysis of Non Markov Systems
Char for Network Archtectures ad Servces Prof. Carle Departmet of Computer Scece U Müche Aalyss of System Performace IN2072 Chapter 5 Aalyss of No Markov Systems Dr. Alexader Kle Prof. Dr.-Ig. Georg Carle
More informationBayes Interval Estimation for binomial proportion and difference of two binomial proportions with Simulation Study
IJIEST Iteratoal Joural of Iovatve Scece, Egeerg & Techology, Vol. Issue 5, July 04. Bayes Iterval Estmato for bomal proporto ad dfferece of two bomal proportos wth Smulato Study Masoud Gaj, Solmaz hlmad
More informationThe number of observed cases The number of parameters. ith case of the dichotomous dependent variable. the ith case of the jth parameter
LOGISTIC REGRESSION Notato Model Logstc regresso regresses a dchotomous depedet varable o a set of depedet varables. Several methods are mplemeted for selectg the depedet varables. The followg otato s
More informationLecture 7. Confidence Intervals and Hypothesis Tests in the Simple CLR Model
Lecture 7. Cofdece Itervals ad Hypothess Tests the Smple CLR Model I lecture 6 we troduced the Classcal Lear Regresso (CLR) model that s the radom expermet of whch the data Y,,, K, are the outcomes. The
More informationAnalysis of a Repairable (n-1)-out-of-n: G System with Failure and Repair Times Arbitrarily Distributed
Amerca Joural of Mathematcs ad Statstcs. ; (: -8 DOI:.593/j.ajms.. Aalyss of a Reparable (--out-of-: G System wth Falure ad Repar Tmes Arbtrarly Dstrbuted M. Gherda, M. Boushaba, Departmet of Mathematcs,
More information{ }{ ( )} (, ) = ( ) ( ) ( ) Chapter 14 Exercises in Sampling Theory. Exercise 1 (Simple random sampling): Solution:
Chapter 4 Exercses Samplg Theory Exercse (Smple radom samplg: Let there be two correlated radom varables X ad A sample of sze s draw from a populato by smple radom samplg wthout replacemet The observed
More informationChapter 8. Inferences about More Than Two Population Central Values
Chapter 8. Ifereces about More Tha Two Populato Cetral Values Case tudy: Effect of Tmg of the Treatmet of Port-We tas wth Lasers ) To vestgate whether treatmet at a youg age would yeld better results tha
More informationCHAPTER 4 RADICAL EXPRESSIONS
6 CHAPTER RADICAL EXPRESSIONS. The th Root of a Real Number A real umber a s called the th root of a real umber b f Thus, for example: s a square root of sce. s also a square root of sce ( ). s a cube
More informationSimulation Output Analysis
Smulato Output Aalyss Summary Examples Parameter Estmato Sample Mea ad Varace Pot ad Iterval Estmato ermatg ad o-ermatg Smulato Mea Square Errors Example: Sgle Server Queueg System x(t) S 4 S 4 S 3 S 5
More informationDiscrete Mathematics and Probability Theory Fall 2016 Seshia and Walrand DIS 10b
CS 70 Dscrete Mathematcs ad Probablty Theory Fall 206 Sesha ad Walrad DIS 0b. Wll I Get My Package? Seaky delvery guy of some compay s out delverg packages to customers. Not oly does he had a radom package
More information8.1 Hashing Algorithms
CS787: Advaced Algorthms Scrbe: Mayak Maheshwar, Chrs Hrchs Lecturer: Shuch Chawla Topc: Hashg ad NP-Completeess Date: September 21 2007 Prevously we looked at applcatos of radomzed algorthms, ad bega
More informationUnimodality Tests for Global Optimization of Single Variable Functions Using Statistical Methods
Malaysa Umodalty Joural Tests of Mathematcal for Global Optmzato Sceces (): of 05 Sgle - 5 Varable (007) Fuctos Usg Statstcal Methods Umodalty Tests for Global Optmzato of Sgle Varable Fuctos Usg Statstcal
More informationTHE ROYAL STATISTICAL SOCIETY HIGHER CERTIFICATE
THE ROYAL STATISTICAL SOCIETY 00 EXAMINATIONS SOLUTIONS HIGHER CERTIFICATE PAPER I STATISTICAL THEORY The Socety provdes these solutos to assst caddates preparg for the examatos future years ad for the
More informationDimensionality Reduction and Learning
CMSC 35900 (Sprg 009) Large Scale Learg Lecture: 3 Dmesoalty Reducto ad Learg Istructors: Sham Kakade ad Greg Shakharovch L Supervsed Methods ad Dmesoalty Reducto The theme of these two lectures s that
More informationLecture 3 Probability review (cont d)
STATS 00: Itroducto to Statstcal Iferece Autum 06 Lecture 3 Probablty revew (cot d) 3. Jot dstrbutos If radom varables X,..., X k are depedet, the ther dstrbuto may be specfed by specfyg the dvdual dstrbuto
More informationABOUT ONE APPROACH TO APPROXIMATION OF CONTINUOUS FUNCTION BY THREE-LAYERED NEURAL NETWORK
ABOUT ONE APPROACH TO APPROXIMATION OF CONTINUOUS FUNCTION BY THREE-LAYERED NEURAL NETWORK Ram Rzayev Cyberetc Isttute of the Natoal Scece Academy of Azerbaa Republc ramrza@yahoo.com Aygu Alasgarova Khazar
More informationArithmetic Mean and Geometric Mean
Acta Mathematca Ntresa Vol, No, p 43 48 ISSN 453-6083 Arthmetc Mea ad Geometrc Mea Mare Varga a * Peter Mchalča b a Departmet of Mathematcs, Faculty of Natural Sceces, Costate the Phlosopher Uversty Ntra,
More informationTHE ROYAL STATISTICAL SOCIETY GRADUATE DIPLOMA
THE ROYAL STATISTICAL SOCIETY EXAMINATIONS SOLUTIONS GRADUATE DIPLOMA PAPER II STATISTICAL THEORY & METHODS The Socety provdes these solutos to assst caddates preparg for the examatos future years ad for
More informationMultiple Regression. More than 2 variables! Grade on Final. Multiple Regression 11/21/2012. Exam 2 Grades. Exam 2 Re-grades
STAT 101 Dr. Kar Lock Morga 11/20/12 Exam 2 Grades Multple Regresso SECTIONS 9.2, 10.1, 10.2 Multple explaatory varables (10.1) Parttog varablty R 2, ANOVA (9.2) Codtos resdual plot (10.2) Trasformatos
More informationChapter 13 Student Lecture Notes 13-1
Chapter 3 Studet Lecture Notes 3- Basc Busess Statstcs (9 th Edto) Chapter 3 Smple Lear Regresso 4 Pretce-Hall, Ic. Chap 3- Chapter Topcs Types of Regresso Models Determg the Smple Lear Regresso Equato
More informationInvestigation of Partially Conditional RP Model with Response Error. Ed Stanek
Partally Codtoal Radom Permutato Model 7- vestgato of Partally Codtoal RP Model wth Respose Error TRODUCTO Ed Staek We explore the predctor that wll result a smple radom sample wth respose error whe a
More informationDimensionality reduction Feature selection
CS 750 Mache Learg Lecture 3 Dmesoalty reducto Feature selecto Mlos Hauskrecht mlos@cs.ptt.edu 539 Seott Square CS 750 Mache Learg Dmesoalty reducto. Motvato. Classfcato problem eample: We have a put data
More informationUNIVERSITY OF OSLO DEPARTMENT OF ECONOMICS
UNIVERSITY OF OSLO DEPARTMENT OF ECONOMICS Postpoed exam: ECON430 Statstcs Date of exam: Jauary 0, 0 Tme for exam: 09:00 a.m. :00 oo The problem set covers 5 pages Resources allowed: All wrtte ad prted
More informationLogistic regression (continued)
STAT562 page 138 Logstc regresso (cotued) Suppose we ow cosder more complex models to descrbe the relatoshp betwee a categorcal respose varable (Y) that takes o two (2) possble outcomes ad a set of p explaatory
More informationDescriptive Statistics
Page Techcal Math II Descrptve Statstcs Descrptve Statstcs Descrptve statstcs s the body of methods used to represet ad summarze sets of data. A descrpto of how a set of measuremets (for eample, people
More informationCubic Nonpolynomial Spline Approach to the Solution of a Second Order Two-Point Boundary Value Problem
Joural of Amerca Scece ;6( Cubc Nopolyomal Sple Approach to the Soluto of a Secod Order Two-Pot Boudary Value Problem W.K. Zahra, F.A. Abd El-Salam, A.A. El-Sabbagh ad Z.A. ZAk * Departmet of Egeerg athematcs
More informationChapter 14 Logistic Regression Models
Chapter 4 Logstc Regresso Models I the lear regresso model X β + ε, there are two types of varables explaatory varables X, X,, X k ad study varable y These varables ca be measured o a cotuous scale as
More informationLecture Notes Types of economic variables
Lecture Notes 3 1. Types of ecoomc varables () Cotuous varable takes o a cotuum the sample space, such as all pots o a le or all real umbers Example: GDP, Polluto cocetrato, etc. () Dscrete varables fte
More informationRuntime analysis RLS on OneMax. Heuristic Optimization
Lecture 6 Rutme aalyss RLS o OeMax trals of {,, },, l ( + ɛ) l ( ɛ)( ) l Algorthm Egeerg Group Hasso Platter Isttute, Uversty of Potsdam 9 May T, We wat to rgorously uderstad ths behavor 9 May / Rutme
More informationLikewise, properties of the optimal policy for equipment replacement & maintenance problems can be used to reduce the computation.
Whe solvg a vetory repleshmet problem usg a MDP model, kowg that the optmal polcy s of the form (s,s) ca reduce the computatoal burde. That s, f t s optmal to replesh the vetory whe the vetory level s,
More informationKLT Tracker. Alignment. 1. Detect Harris corners in the first frame. 2. For each Harris corner compute motion between consecutive frames
KLT Tracker Tracker. Detect Harrs corers the frst frame 2. For each Harrs corer compute moto betwee cosecutve frames (Algmet). 3. Lk moto vectors successve frames to get a track 4. Itroduce ew Harrs pots
More informationLecture 2 - What are component and system reliability and how it can be improved?
Lecture 2 - What are compoet ad system relablty ad how t ca be mproved? Relablty s a measure of the qualty of the product over the log ru. The cocept of relablty s a exteded tme perod over whch the expected
More informationhp calculators HP 30S Statistics Averages and Standard Deviations Average and Standard Deviation Practice Finding Averages and Standard Deviations
HP 30S Statstcs Averages ad Stadard Devatos Average ad Stadard Devato Practce Fdg Averages ad Stadard Devatos HP 30S Statstcs Averages ad Stadard Devatos Average ad stadard devato The HP 30S provdes several
More informationFault Diagnosis Using Feature Vectors and Fuzzy Fault Pattern Rulebase
Fault Dagoss Usg Feature Vectors ad Fuzzy Fault Patter Rulebase Prepared by: FL Lews Updated: Wedesday, ovember 03, 004 Feature Vectors The requred puts for the dagostc models are termed the feature vectors
More informationMAX-MIN AND MIN-MAX VALUES OF VARIOUS MEASURES OF FUZZY DIVERGENCE
merca Jr of Mathematcs ad Sceces Vol, No,(Jauary 0) Copyrght Md Reader Publcatos wwwjouralshubcom MX-MIN ND MIN-MX VLUES OF VRIOUS MESURES OF FUZZY DIVERGENCE RKTul Departmet of Mathematcs SSM College
More information3. Basic Concepts: Consequences and Properties
: 3. Basc Cocepts: Cosequeces ad Propertes Markku Jutt Overvew More advaced cosequeces ad propertes of the basc cocepts troduced the prevous lecture are derved. Source The materal s maly based o Sectos.6.8
More informationSimple Linear Regression
Statstcal Methods I (EST 75) Page 139 Smple Lear Regresso Smple regresso applcatos are used to ft a model descrbg a lear relatoshp betwee two varables. The aspects of least squares regresso ad correlato
More informationSupervised learning: Linear regression Logistic regression
CS 57 Itroducto to AI Lecture 4 Supervsed learg: Lear regresso Logstc regresso Mlos Hauskrecht mlos@cs.ptt.edu 539 Seott Square CS 57 Itro to AI Data: D { D D.. D D Supervsed learg d a set of eamples s
More informationENGI 4421 Joint Probability Distributions Page Joint Probability Distributions [Navidi sections 2.5 and 2.6; Devore sections
ENGI 441 Jot Probablty Dstrbutos Page 7-01 Jot Probablty Dstrbutos [Navd sectos.5 ad.6; Devore sectos 5.1-5.] The jot probablty mass fucto of two dscrete radom quattes, s, P ad p x y x y The margal probablty
More informationBeam Warming Second-Order Upwind Method
Beam Warmg Secod-Order Upwd Method Petr Valeta Jauary 6, 015 Ths documet s a part of the assessmet work for the subject 1DRP Dfferetal Equatos o Computer lectured o FNSPE CTU Prague. Abstract Ths documet
More informationNaïve Bayes MIT Course Notes Cynthia Rudin
Thaks to Şeyda Ertek Credt: Ng, Mtchell Naïve Bayes MIT 5.097 Course Notes Cytha Rud The Naïve Bayes algorthm comes from a geeratve model. There s a mportat dstcto betwee geeratve ad dscrmatve models.
More informationBootstrap Method for Testing of Equality of Several Coefficients of Variation
Cloud Publcatos Iteratoal Joural of Advaced Mathematcs ad Statstcs Volume, pp. -6, Artcle ID Sc- Research Artcle Ope Access Bootstrap Method for Testg of Equalty of Several Coeffcets of Varato Dr. Navee
More informationObjectives of Multiple Regression
Obectves of Multple Regresso Establsh the lear equato that best predcts values of a depedet varable Y usg more tha oe eplaator varable from a large set of potetal predctors {,,... k }. Fd that subset of
More informationCS286.2 Lecture 4: Dinur s Proof of the PCP Theorem
CS86. Lecture 4: Dur s Proof of the PCP Theorem Scrbe: Thom Bohdaowcz Prevously, we have prove a weak verso of the PCP theorem: NP PCP 1,1/ (r = poly, q = O(1)). Wth ths result we have the desred costat
More informationVOL. 3, NO. 11, November 2013 ISSN ARPN Journal of Science and Technology All rights reserved.
VOL., NO., November 0 ISSN 5-77 ARPN Joural of Scece ad Techology 0-0. All rghts reserved. http://www.ejouralofscece.org Usg Square-Root Iverted Gamma Dstrbuto as Pror to Draw Iferece o the Raylegh Dstrbuto
More informationModule 7: Probability and Statistics
Lecture 4: Goodess of ft tests. Itroducto Module 7: Probablty ad Statstcs I the prevous two lectures, the cocepts, steps ad applcatos of Hypotheses testg were dscussed. Hypotheses testg may be used to
More informationMULTIDIMENSIONAL HETEROGENEOUS VARIABLE PREDICTION BASED ON EXPERTS STATEMENTS. Gennadiy Lbov, Maxim Gerasimov
Iteratoal Boo Seres "Iformato Scece ad Computg" 97 MULTIIMNSIONAL HTROGNOUS VARIABL PRICTION BAS ON PRTS STATMNTS Geady Lbov Maxm Gerasmov Abstract: I the wors [ ] we proposed a approach of formg a cosesus
More information