Parameter Estimation for Relational Kalman Filtering
|
|
- Audra Nichols
- 6 years ago
- Views:
Transcription
1 Sascal Relaonal AI: Papers from he AAAI-4 Workshop Parameer Esmaon for Relaonal Kalman Flerng Jaesk Cho School of Elecrcal and Compuer Engneerng Ulsan Naonal Insue of Scence and Technology Ulsan, Korea Eyal Amr Deparmen of Compuer Scence Unversy of Illnos a Urbana-Champagn Urbana, IL, USA Tanfang Xu and Alber J Valocch Deparmen of Cvl and Envronmenal Engneerng Unversy of Illnos a Urbana-Champagn Urbana, IL, USA Absrac The Kalman Fler (KF) s pervasvely used o conrol a vas array of consumer, healh and defense producs By groupng ses of symmerc sae varables, he Relaonal Kalman Fler (RKF) enables o scale he exac KF for large-scale dynamc sysems In hs paper, we provde a parameer learnng algorhm for RKF, and a regroupng algorhm ha prevens he degeneraon of he relaonal srucure for effcen flerng The proposed algorhms sgnfcanly expand he applcably of he RKFs by solvng he followng quesons: () how o learn parameers for RKF n paral observaons; and () how o regroup he degeneraed sae varables by nosy real-world observaons We show ha our new algorhms mprove he effcency of flerng he largescale dynamc sysem Inroducon any real-world sysems can be modeled by connuous varables and relaonshps (or dependences) among hem The Kalman Fler (KF) (Kalman 960) accuraely esmaes he sae of varables n a lnear dynamc sysem wh Gaussan nose gven a sequence of conrol-npus and observaons The KF has been appled n a broad range of domans such as robocs, fnance (Bahman-Oskooee and Brown 004) and envronmenal scence (FP and Berkens 00; Clark e al 008) Gven a sequence of observaons and lnear dependences wh Gaussan nose beween varables, he KF calculaes he condonal probably densy of he sae varables a each me sep Unforunaely, he KF compuaons are cubc n he number of sae varables, whch lms he use of exsng exac mehods o domans wh a large number of sae varables Ths has led o he combnaon of approxmaon and samplng n he Ensemble Kalman Fler (Evensen 994), and recenly o he Relaonal Kalman Flers (RKFs) over grouped sae varables (Cho, Guzman-Rvera, and Amr 0; Ahmad, Kersng, and Sanner 0) The RKFs Copyrgh c 04, Assocaon for he Advancemen of Arfcal Inellgence (wwwaaaorg) All rghs reserved leverage he ably of relaonal languages o specfy models wh sze of represenaon ndependen of he sze of populaons nvolved (Fredman e al 999; Poole 003; Rchardson and Domngos 006) Lfed nference algorhms for relaonal connuous models (Wang and Domngos 008; Cho, Hll, and Amr 00; Ahmad, Kersng, and Sanner 0) degenerae (or spl) relaonal srucures upon ndvdual observaons Lfed RKF (Cho, Guzman-Rvera, and Amr 0) manans relaonal srucure when he same number of observaons are made Oherwse, also degeneraes (possbly rapdly) he relaonal srucure, hus lfed RKF may no be useful wh sparse observaons The man conrbuons of hs paper are () o regroup he degeneraed sae varables from nosy real-world observaons wh a gh error bounds; and () o learn parameers for RKFs We propose a new learnng algorhm for RKFs We show ha he relaonal learnng expedes flerng, and acheves accurae flerng n heory and pracce The key nuon s ha he axmum Lkelhood Esmae (LE) of RKF parameers s he emprcal mean and varance over sae varables ncluded n each group For paral observaons, he parameers can be calculaed smlarly We show ha under reasonable condons varances of degeneraed sae varables on paral observaons converge exponenally Thus, our approxmae regroupng algorhm has bounded errors compared o he exac KF We could show ha he RKF wh regroupng s more robus agans degeneracy han he Lfed RKF n pracce wh paral observaons Relaonal Lnear Dynamc Sysems In hs secon, we defne relaonal lnear dynamc sysems Dependences among varables are represened usng relaonal aoms, or jus aoms The relaonal aoms are useful when he jon probably of varables nvolves common ypes of funcons When represenng he jon probably For comprehensve defnons, see (Poole 003; de Salvo Braz, Amr, and Roh 005)
2 dsrbuon, here are producs of he parameerzed funcons (or poenals) 03 W (405N,03W) r (40N,03W) 0 W 0 W r 00 W 99 W 98 W 4 N 40 N A parwse Gaussan facor, or jus a facor, f = ((x, x ), φ) s a par where φ s a poenal funcon on (x, x ) from R o R + where (x, x ) s a par of ground random varables derved by ground subsuons from (X(Lθ), X (L θ )) A facor f defnes a weghng funcon on a valuaon (x, x ) = (v, v ): w f (x, x ) = φ(v, v ) The weghng funcon for a parfacor g s he produc of he weghng funcons over all of s ground subsuons (facors), w g (v) = f g w f (v) Hence, a se of parfacors G defnes a probably densy, w G (v) = w f (v), Z g G f g Colorado Nebraska Kansas les Sream nework 39 N where Z s he normalzng consan 3 In hs way, we can represen he jon probably of all random varables (eg all wells n regon and regon ) Relaonal Transon odels (RTs) characerze he dependence of relaonal aoms beween me seps X (a) and X j + (a ) are relaonal aoms a me sep and + respecvely when a and a are ground subsuons, eg, θ =(40N,98W) U (a) s he conrol-npu nformaon A RT akes he followng form, Fgure : Republcan Rver Basn coverng porons of easern Colorado, norhwes Kansas and souhwes Nebraska Ths fgure shows wo clusered waer wells; regon (r ) and regon (r ) Waer wells n each regon have he same (lnear Gaussan) relaonshps wh wells n oher regons Relaonal aoms represen he se of sae varables correspondng o all ground subsuons of s parameer varables For example, le X r (Laude, Longude) be an aom for he (waer level of) wells n regon, θ=(40n,03w ) When we subsue Laude and Longude wh θ, he aom becomes a sae varable X r (40N, 03W ) whch represens he level (or predcon) of well head a (Laude=40N, Longude=03W ) Formally, applyng a subsuon θ o an aom X(L) yelds a new aom X(Lθ) where Lθ s obaned by renamng he parameer varables n L accordng o θ If θ s a ground subsuon, X(Lθ) s a ground sae varable lke X r (40N, 03W ) X(L) or jus X denoes he he number of dsnc sae varables generaed from X by all subsuons A parwse Gaussan parfacor ((X, X ), φ) s composed of a par of wo aoms (X, X ) and a lnear Gaussan poenal φ beween wo aoms n he followng form [ (X X φ(x, X µ) ] ) exp For example, a parwse Gaussan parfacor φ r,r (X r, X r ) represens he lnear Gaussan relaonshp beween wo ground varables chosen from regon and regon respecvely Here, we assume ha he ground sae varables are unvarae, eg, doman of x s R However, s sraghforward o exend hs model o he one wh mulvarae ground varables X j + (a ) = B j X X (a) + B j U U (a) + G j RT, () where G j RT N (0, j RT ) and N (m, ) s he normal dsrbuon wh mean m and varance B j X and Bj U are he lnear ranson coeffcens In he lnear Gaussan represenaon, he ranson models ake he followng form, φ RT (X j + (a ) X(a), U (a)) [ exp (Xj + (a ) B j X X (a) B j U U ] (a)) () j RT The mos common ranson s he one from he sae X (a) o he sae self X +(a) a he nex me sep, X +(a) = B XX (a) + B U U (a) + G RT (3) Relaonal Observaon odels (s) represen he relaonshps beween he hdden (sae) varables, X (a), and he observaons made drecly on he relaed varable, O (a) (drec observaons), O (a) = H X (a) + G, G N (0, ) (4) H s he lnear coeffcen from X (a) o O (a) s also represen he relaonshps beween he hdden varables X (a) and he observaons made ndrecly on a varable n he aom O (a ) where a a (relaonal observaons), O (a ) = H X (a)+g, G N (0, ) (5) 3 The codon of beng a probably densy s ha a leas a random varable has a pror dsrbuon, see (Cho, Hll, and Amr 00)
3 n mos cases, s reasonable o se he varance of drec observaon var(g ) s smaller han he varance of relaonal one var(g ) s For he well example, X r (40N, 03W ), an observaon made a he exac locaon O r (40N, 03W ) wll have he smaller varance (more ceran) han an observaon made a a nearby locaon O r (405N, 03W ) In he lnear Gaussan represenaon, s ake he followng form, [ ] φ (O(a) X (a)) exp (O (a) HX (a)) Relaonal Parwse odels (RPs) represen lnear dependences beween pars of relaonal aoms, R j X(a) = R j X j (a ) + G j RP, Gj RP N (0, j RP ) (6) s he coeffcen Noe ha RTs and s represen he naure of dynamc sysems (eg he sae a he nex me sep depends on he curren me sep) The produc of RPs s an effcen way o represen he relaonal srucure over aoms, groups of sae varables Relaonal Kalman Fler (RKF) s a flerng procedure wh a relaonal lnear dynamc sysem whch s composed of RTs, s and RPs Tha s, he jon probably of sae varables are represened by he produc of parwse Gaussan parfacors Lfed RKF compues he poseror of he sae varables gven a pror (curren) belef and full or paral observaons The npu o he problem s: () relaonal parfacors (RTs, RPs and s); () curren belef over aoms (X0); (3) a sequence of conrolnpus (U,, UT ); and (4) a sequence of observaons (O,, OT ) The oupu s he mulvarae Gaussan dsrbuon over he aoms (XT ) a each me sep T The flerng problem s solved by algorhms represened (Cho, Guzman-Rvera, and Amr 0; Ahmad, Kersng, and Sanner 0) Here, we urn our focus o he parameer learnng problem 3 Learnng Relaonal Kalman Fler The wo mporan parameers of he RKF are he ranson models and observaon models In hs secon, we presen a learnng algorhm ha derves he maxmum lkelhood esmaes of RTs and s For smplcy, we wll presen a soluon wh fully observed model, frs A soluon for paral observaons can be derved wh a slgh modfcaon 3 Algorhm LearnngRKF Algorhm LearnngRKF esmaes he parameer of RKF gven a sequence of observaons such as measuremens of waer wells for several years The overall procedure s smlar o he one wh he parameer learnng for he ground KF Here, he man dfference s ha he coeffcens and covarances of RTs and s are he block marces A subroune, BlockAverage, averages ou he dagonal and nondagonal enres of an npu marx, and hen oupus a block marx where each block ncludes he emprcal means, varances and covarances n each block In he followng secons, we wll show ha he block marx compued by BlockAverage s he LE esmae Algorhm LearnngRKF npu: a sequence of obs (O,, O T ) (B, Σ T, C, Σ O ) (I, I, I, I) currenll repea prevll currenll (B, Σ T, C, Σ O) LearnGroundT(O,B,Σ T,C,Σ O) (B, Σ T, C, Σ O ) BlockAverage(B, Σ T, C, Σ O ) currenll log P (O X, B, Σ T, C, Σ O ) unl prevll - currenll < ɛ oupu: esmaed parameers (B, Σ T, C, Σ O ) 3 Learnng Transon odels Here, we derve he parameer of he RTs: lnear coeffcen B and Gaussan nose G RT Learnng Transon Nose s o calculae he mean and he covarance marx n he followng block forms, Σ, Σ, Σ,n µ T = µ µ µ n, Σ Σ, Σ, Σ,n T = (7) Σ n, Σ n, Σ n,n Where µ s a vecor of sze X (=n ); Σ,j s a marx of sze n by n j Gven a pror, a lnear coeffcen B and a sequence of full observaons, we derve he esmae X a me sep assumng he Gaussan nose n he ranson model The LE esmaon of µ and Σ for he RT can be derved: (µ T max, Σ T max ) = arg max log f N ( X ; µ T, Σ T ) µ T,Σ T =,,T where X =X BX Proposon Gven a RKF wh a sngle aom, he maxmum lkelhood esmaes of he Gaussan ranson nose are he emprcal mean, varance and covarance as follows, m m µ LE = such ha, m= n T m T = a A = n(n ) T, Σ LE = T = X (a), = n T T ( X (a) m) = a A ( ) ( X (a) m X (a ) m) a,a A a a where n = X (A) and T =T 4
4 Proof The LEs of he parameers (µ T, Σ T ) are derved by he paral dervaves of he log lkelhood, µ T Σ T T log f N ( X ; µ T, Σ T )=0, = T log f N ( X ; µ T, Σ T )=0 = All ground varables generaed from he aom X have he same mean, varance and covarances as shown n Equaon (8) Now, we can specfy he followng lnear consrans; m = T X (a ) = = T X (a n ) Tha s, m= n T a X (a) The covarance marx of he RT s also calculaed from he emprcal covarance marx The dagonal enres are derved for he varances; = T ( X (a ) m) = = T ( X (a n ) m) Thus, = ( X n T a (a) m) Non-dagonal enres (covarances) are derved smlarly wh n(n ) emprcal covarances 4 Ths resul s conssen wh he resul n non-relaonal KF because he LE esmaes of he ground KF (µ T and Σ T ) are known o be he emprcal mean and he emprcal covarance marx (Rowes and Ghahraman 999) In he general case, when we have mulple aoms, he mean vecor and he covarance marx are block forms as shown n Equaon 7 Tha s, he mean and covarance values are same n each subblock In case of, wo aoms X and X j, he means and covarances are as follows: µ T = [ ] µ, Σ µ T = [ ] Σ, Σ, Σ, Σ, The LE parameers of he RT are derved smlarly wh emprcal means and covarances of subblocks Proposon Gven a RKF wh mulple aoms, he maxmum lkelhood esmaes of he Gaussan ranson nose are he emprcal means, varances and covarances, µ = [m,, m ] T s m = n T T X (a), = Σ, = Σ,j = 4 One ges he same resul when dfferenang m, and drecly from he log-lkelhood a = n T = = T ( ) ( ) X (a) m X (a) m = a n (n ) T n n j T T T = a,a A (a a ) = a A,b B where n = X and T =T ( X (a) m ) ( X (a ) m ), ( X (a) m ) ( X j (b) m j ) Proof The prncples used n he proof of Proposon are appled because he Σ, and Σ,j are block marces Learnng Lnear Coeffcen s o esmae he lnear coeffcen B beween X and X In hs case, gven oher parameers, Gaussan nose of RTs and s, he LE of B s derved as follows (Rowes and Ghahraman 999), B = X X T X X T =,,T =,,T Here, we call he B lnear coeffcen of he ground T, and B s reshaped as a block marx by averagng he coeffcen n each subblock When B,j denoes he subblock for he lnear ranson from X j o X The LE of he block coeffcen s represened as follows, b b b b b b b b B, = B,j = b b b b b such ha, b= n, n k= B k,k, b = n (n ) b = (n,n j),j n n j (k,l)=(,) B k,l (n,n ) (k,l)=(,) B, k,l, k l BlockAverage n Algorhm LearnngRKF denoes hs averagng-ou procedure The block coeffcen marx B s also he LE of RT 33 Learnng Observaon odels Gven RTs and a sequence of full observaons, we derve he esmae X a me sep assumng ha here s no observaon nose Learnng Observaon Nose s o esmae he mean vecor and covarance marx for he The LEs problem s formulaed as follows, (µ LE, Σ LE ) = arg max µ O,Σ O T log N ( O ; µ O, Σ O ) = where O = O C X The dervaon s smlar o RTs One can subsue O for X n Proposon 5
5 Learnng Lnear Coeffcen C s o calculae he lnear coeffcen beween X and O Ĉ = O X T X X T =,,T =,,T Here C s also calculaed from Ĉ by averagng ou each subblock as n learnng B 4 LRKF wh Regroupngs Wh he esmaed parameers, he RKF predcs he sae varables n he relaonal lnear dynamc models Ths secon presens a new lfed Kalman flerng algorhm, whch approxmaely regroups degeneraed relaonal srucures Exsng lfed Kalman flerng algorhms (Cho, Guzman- Rvera, and Amr 0; Ahmad, Kersng, and Sanner 0) suffer degeneraons of relaonal srucures when sparse observaons are made 5 Algorhm LRKF-Regroup also degeneraes he domans of relaonal aoms by callng DegenAom when sae varables are observed n dfferen me, hus dfferen Obs (eg, Obs (,a) Obs (,a )) Here, Obs (,a) sores he mos recenly observed me for a ground subson a n he -h aom To overcome such degeneracy, LRKF-Regroup nroduces a new subroune, called ergeaom, whch merges covarance srucures when random varables are no drecly observed for a ceran me seps, say k Algorhm LRKF-Regroup (Predcon w/ esng daa) Inpu: params (B, Σ T, C, Σ O ), obs (O,, O T ) repea µ 0 0, Σ 0 0 (Obs (,),,Obs (n,n)) (0,, 0) for o T do (µ, Σ ) Predc-RT(µ, Σ, B, Σ T ) for all (, a) s O (a) s observed do Obs (,a) (B,Σ T,C,Σ O ) DegenAom(Obs, B,Σ T, C, Σ O ) (µ, Σ ) Updae-(µ, Σ, C, Σ O ) (B,Σ T,C,Σ O ) ergeaom(obs,, B,Σ T,C,Σ O ) unl s T Oupu: sae esmaons ((µ 0, Σ 0 ),, (µ T, Σ T )) The ergeaom operaon eraes aoms and fnd all sae varables whch are no observed for a ceran me seps k The seleced varables are sored n mls Then, he Blockerge respecvely averages dagonal enres and nondagonal enres n mls, and ses he averaged values o sae varables n mls In hs way, rebulds he compac relaonal srucure agan 5 Noe ha, he lfed algorhm n (Cho, Guzman-Rvera, and Amr 0) only vald when he same number of observaons are made a he same me seps Algorhm 3 ergeaom npu: recen obs me Obs, me, params (B,Σ T,C,Σ O ) mls for = o n do for each a s Obs (,a) +k do mls mls {a} (B,Σ T, C, Σ O ) = BlockAverage(mls, B, Σ T, C, Σ O ) oupu: merged relaonal srucures (B, Σ T, C, Σ O ) Lemma 3 When () a leas one relaonal observaon s made on O a each me sep; he varance of any sae varable X (a) n LRKF-Regroup s bounded by and converges o RT RT Proof Le he varance of he -h aom be RT and he varances of drec and relaonal observaons respecvely be and where < as Equaon (4) and (5) Le (a) be he varance of a sae varable X (a) a me Ths varable wll be updaed by a leas one relaonal observaon by he LRKF-Regroup Then, he new varance s = + + (a) (a) Tha s, +(a) mn((a), ) Use he followng equaon for ranson and updae n each flerng sep, + (a) (a) + RT + For he convergence, le +(a)= (a)= (a) The varance s (a) = RT RT Theorem 4 When () a leas one relaonal observaon s made on O a each me sep; and () no drec observaon s made on wo sae varables X (a) and X (a ) a leas for k me seps, he dfference of he varances of wo sae varables (a) and (a ) s bounded by c k RT (+ (a) / ) where c s (a) /( (a) + ) and c Proof We follow he resul of he Lemma 3 and use (a) The varance of each me sep follows he recursve form, + = (a) (a) + RT + (8) An exac (non-recursve) formula for +(a) s non rval Thus, we nroduce anoher smpler, convergng sequence, +(a) = c( (a) + RT ) Snce (a) (a) s posve and convex when c <, 0 (a) (a) (a), (a) (a) (a) (a) 6
6 Varance w/o relaonal obs w/ relaonal obs In he synhec daase, we assume ha an aom wh 300 ground subsuons, X =30 Then we make a sparse observaons wh a rae of 90% Tha s, 90% of sae varables wll be observed n each me sep Then, we repor he numbers of shaered (or degeneraed) groups and average flerng me n he Lfed RKF and n our LRKF-Regroup The resul s shown n Fgure Tme seps Fgure : Varances of sae varables wh one observaon per 0 me seps whou a relaonal obs (square shaped marks) and wh relaonal obs (crcle shaped marks) Ths fgure shows he varances (y axs) n me seps (x axs) Seng ha RT =0, =0, =05, and le observaons made a dfferen me seps Ths smulaon shows he nuon of Lemma 3 The convergence of he smpler form s slower han he orgnal one n Equaon (8) However, provdes an exac formulaon and converge exponenally, (a) = c k k(a) + RT c k c WLOG, we se X (a ) has no drec observaon longer han X (a) The varance of X (a ) has he same formulaon wh a subsuon of k+α for k Thus, varance of X (a ) s (a ) =c k+α k α(a ) +RT c k+α c Noe ha, +α(a) +α(a ) + (a ) +α(a ) (a) + (a ) (a) = c k ( RT /( c) +k (a) ) = c k ( RT ( + (a) / ) +k (a) ) c k RT ( + (a) / ) Fgure 3: Comparons of Lfed RKF and LRKF-Regoup n a smulaon daase To verfy he effcency of Lfed-Regroup, we use a regonal groundwaer flow ODFLOW model, he Republcan Rver Compac Assocaon (RRCA) odel (ckusck 003) The exraced daase s a small subse of he whole ground waer model, monhly measured head (waer level) a over 3,000 wells We choose a closely relaed 00 wells and randomly selece 00 measuremen perods (00 monhs) Then, we repor he degeneraces n Lfed-RKF and LRKF- Regroup and average flerng me affeced by he degeneraons n Fgure 4 5 Expermenal Resuls To compare he mprove he effcency, we mplemen he Lfed RKF (Cho, Guzman-Rvera, and Amr 0) and our LRKF-Regroup The algorhm s compared n wo daases wh sparse observaons: one synhec daase and one realworld ground waer daase Noe ha, he lfed RKF wll no degenerae model on full, dense, observaons In boh expermen, we se k o be 4 Tha s, wo sae varables wll be merged f hey have he same observaon numbers and ypes when a leas a relaonal observaon s made Wh relaonal observaons, he varances of degeneraed sae varables are reasonably small even afer 4 me seps Fgure 4: Comparons of Lfed RKF and LRKF-Regoup n he groundwaer model 7
7 6 Concluson Ths paper provdes new answers and nsghs on () how o learn parameers for RKF; and () how o regroup he sae varables from nosy real-world daa We propose a new algorhm ha regroups he sae varables when ndvdual observaons are made o RKF n dfferen me seps We use he RKF n a smulaed and a real-world daase, and demonsrae ha he RKF mproves he effcency of flerng he large-scale dynamc sysem 7 Acknowledgmens Ths work s suppored by he Naonal Scence Foundaon Hydrologc Scence Program under Gran No Ths wor was suppored by he year of 04 Research Fund of UNIST (Ulsan Naonal Insue of Scence and Technology) ckusck, V 003 Fnal repor for he specal maser wh cerfcae of adopon of rrca groundwaer model Sae of Kansas v Sae of Nebraska and Sae of Colorado, n he Supreme Cour of he Uned Saes 3605(45):4440 Poole, D 003 Frs-order probablsc nference In Proceedngs of he Inernaonal Jon Conference on Arfcal Inellgence, Rchardson,, and Domngos, P 006 arkov logc neworks achne Learnng 6(-):07 36 Rowes, S, and Ghahraman, Z 999 A unfyng revew of lnear gaussan models Neural Compu (): Wang, J, and Domngos, P 008 Hybrd markov logc neworks In Proceedngs of he AAAI Conference on Arfcal Inellgence, 06 References Ahmad, B; Kersng, K; and Sanner, S 0 ulevdence lfed message passng, wh applcaon o pagerank and he kalman fler In Proceedngs of he Inernaonal Jon Conference on Arfcal Inellgence Bahman-Oskooee,, and Brown, F 004 Kalman fler approach o esmae he demand for nernaonal reserves Appled Economcs 36(5): Cho, J; Guzman-Rvera, A; and Amr, E 0 Lfed relaonal kalman flerng In Proceedngs of he Inernaonal Jon Conference on Arfcal Inellgence Cho, J; Hll, D J; and Amr, E 00 Lfed nference for relaonal connuous models In Proceedngs of he Conference on Uncerany n Arfcal Inellgence, 6 34 Clark, P; Rupp, D E; Woods, R A; Zheng, X; Ibb, R P; Slaer, A G; Schmd, J; and Uddsrom, J 008 Hydrologcal daa assmlaon wh he ensemble kalman fler: Use of sreamflow observaons o updae saes n a dsrbued hydrologcal model Advances n Waer Resources 3(0): de Salvo Braz, R; Amr, E; and Roh, D 005 Lfed frsorder probablsc nference In Proceedngs of he Inernaonal Jon Conference on Arfcal Inellgence, Evensen, G 994 Sequenal daa assmlaon wh a nonlnear quas-geosrophc model usng mone carlo mehods o forecas error sascs Journal of Geophyscal Research 99: FP,, and Berkens 00 Spao-emporal modellng of he sol waer balance usng a sochasc model and sol profle descrpons Geoderma 03( ):7 50 Fredman, N; Geoor, L; Koller, D; and Pfeffer, A 999 Learnng probablsc relaonal models In Proceedngs of he Inernaonal Jon Conference on Arfcal Inellgence, Kalman, R E 960 A new approach o lnear flerng and predcon problems Transacons of he ASE Journal of Basc Engneerng 8(Seres D):
Fall 2010 Graduate Course on Dynamic Learning
Fall 200 Graduae Course on Dynamc Learnng Chaper 4: Parcle Flers Sepember 27, 200 Byoung-Tak Zhang School of Compuer Scence and Engneerng & Cognve Scence and Bran Scence Programs Seoul aonal Unversy hp://b.snu.ac.kr/~bzhang/
More informationLifted Relational Kalman Filtering
Lfed Relaonal Kalman Flerng Jaesk Cho and Abner Guzman-Rvera and Eyal Amr Compuer Scence Deparmen Unversy of Illnos a Urbana-Champagn Urbana, IL, 61801, USA {jaesk,aguzman5,eyal}@llnos.edu Absrac Kalman
More informationOutline. Probabilistic Model Learning. Probabilistic Model Learning. Probabilistic Model for Time-series Data: Hidden Markov Model
Probablsc Model for Tme-seres Daa: Hdden Markov Model Hrosh Mamsuka Bonformacs Cener Kyoo Unversy Oulne Three Problems for probablsc models n machne learnng. Compung lkelhood 2. Learnng 3. Parsng (predcon
More informationCS434a/541a: Pattern Recognition Prof. Olga Veksler. Lecture 4
CS434a/54a: Paern Recognon Prof. Olga Veksler Lecure 4 Oulne Normal Random Varable Properes Dscrmnan funcons Why Normal Random Varables? Analycally racable Works well when observaon comes form a corruped
More informationV.Abramov - FURTHER ANALYSIS OF CONFIDENCE INTERVALS FOR LARGE CLIENT/SERVER COMPUTER NETWORKS
R&RATA # Vol.) 8, March FURTHER AALYSIS OF COFIDECE ITERVALS FOR LARGE CLIET/SERVER COMPUTER ETWORKS Vyacheslav Abramov School of Mahemacal Scences, Monash Unversy, Buldng 8, Level 4, Clayon Campus, Wellngon
More informationOn One Analytic Method of. Constructing Program Controls
Appled Mahemacal Scences, Vol. 9, 05, no. 8, 409-407 HIKARI Ld, www.m-hkar.com hp://dx.do.org/0.988/ams.05.54349 On One Analyc Mehod of Consrucng Program Conrols A. N. Kvko, S. V. Chsyakov and Yu. E. Balyna
More informationWiH Wei He
Sysem Idenfcaon of onlnear Sae-Space Space Baery odels WH We He wehe@calce.umd.edu Advsor: Dr. Chaochao Chen Deparmen of echancal Engneerng Unversy of aryland, College Par 1 Unversy of aryland Bacground
More informationJohn Geweke a and Gianni Amisano b a Departments of Economics and Statistics, University of Iowa, USA b European Central Bank, Frankfurt, Germany
Herarchcal Markov Normal Mxure models wh Applcaons o Fnancal Asse Reurns Appendx: Proofs of Theorems and Condonal Poseror Dsrbuons John Geweke a and Gann Amsano b a Deparmens of Economcs and Sascs, Unversy
More informationSolution in semi infinite diffusion couples (error function analysis)
Soluon n sem nfne dffuson couples (error funcon analyss) Le us consder now he sem nfne dffuson couple of wo blocks wh concenraon of and I means ha, n a A- bnary sysem, s bondng beween wo blocks made of
More information( ) () we define the interaction representation by the unitary transformation () = ()
Hgher Order Perurbaon Theory Mchael Fowler 3/7/6 The neracon Represenaon Recall ha n he frs par of hs course sequence, we dscussed he chrödnger and Hesenberg represenaons of quanum mechancs here n he chrödnger
More informationFTCS Solution to the Heat Equation
FTCS Soluon o he Hea Equaon ME 448/548 Noes Gerald Reckenwald Porland Sae Unversy Deparmen of Mechancal Engneerng gerry@pdxedu ME 448/548: FTCS Soluon o he Hea Equaon Overvew Use he forward fne d erence
More informationRobustness Experiments with Two Variance Components
Naonal Insue of Sandards and Technology (NIST) Informaon Technology Laboraory (ITL) Sascal Engneerng Dvson (SED) Robusness Expermens wh Two Varance Componens by Ana Ivelsse Avlés avles@ns.gov Conference
More informationIn the complete model, these slopes are ANALYSIS OF VARIANCE FOR THE COMPLETE TWO-WAY MODEL. (! i+1 -! i ) + [(!") i+1,q - [(!
ANALYSIS OF VARIANCE FOR THE COMPLETE TWO-WAY MODEL The frs hng o es n wo-way ANOVA: Is here neracon? "No neracon" means: The man effecs model would f. Ths n urn means: In he neracon plo (wh A on he horzonal
More informationAppendix H: Rarefaction and extrapolation of Hill numbers for incidence data
Anne Chao Ncholas J Goell C seh lzabeh L ander K Ma Rober K Colwell and Aaron M llson 03 Rarefacon and erapolaon wh ll numbers: a framewor for samplng and esmaon n speces dversy sudes cology Monographs
More informationVariants of Pegasos. December 11, 2009
Inroducon Varans of Pegasos SooWoong Ryu bshboy@sanford.edu December, 009 Youngsoo Cho yc344@sanford.edu Developng a new SVM algorhm s ongong research opc. Among many exng SVM algorhms, we wll focus on
More informationLinear Response Theory: The connection between QFT and experiments
Phys540.nb 39 3 Lnear Response Theory: The connecon beween QFT and expermens 3.1. Basc conceps and deas Q: ow do we measure he conducvy of a meal? A: we frs nroduce a weak elecrc feld E, and hen measure
More informationGENERATING CERTAIN QUINTIC IRREDUCIBLE POLYNOMIALS OVER FINITE FIELDS. Youngwoo Ahn and Kitae Kim
Korean J. Mah. 19 (2011), No. 3, pp. 263 272 GENERATING CERTAIN QUINTIC IRREDUCIBLE POLYNOMIALS OVER FINITE FIELDS Youngwoo Ahn and Kae Km Absrac. In he paper [1], an explc correspondence beween ceran
More informationSingle-loop System Reliability-Based Design & Topology Optimization (SRBDO/SRBTO): A Matrix-based System Reliability (MSR) Method
10 h US Naonal Congress on Compuaonal Mechancs Columbus, Oho 16-19, 2009 Sngle-loop Sysem Relably-Based Desgn & Topology Opmzaon (SRBDO/SRBTO): A Marx-based Sysem Relably (MSR) Mehod Tam Nguyen, Junho
More informationChapter 6: AC Circuits
Chaper 6: AC Crcus Chaper 6: Oulne Phasors and he AC Seady Sae AC Crcus A sable, lnear crcu operang n he seady sae wh snusodal excaon (.e., snusodal seady sae. Complee response forced response naural response.
More information[ ] 2. [ ]3 + (Δx i + Δx i 1 ) / 2. Δx i-1 Δx i Δx i+1. TPG4160 Reservoir Simulation 2018 Lecture note 3. page 1 of 5
TPG460 Reservor Smulaon 08 page of 5 DISCRETIZATIO OF THE FOW EQUATIOS As we already have seen, fne dfference appromaons of he paral dervaves appearng n he flow equaons may be obaned from Taylor seres
More informationAn introduction to Support Vector Machine
An nroducon o Suppor Vecor Machne 報告者 : 黃立德 References: Smon Haykn, "Neural Neworks: a comprehensve foundaon, second edon, 999, Chaper 2,6 Nello Chrsann, John Shawe-Tayer, An Inroducon o Suppor Vecor Machnes,
More informationDepartment of Economics University of Toronto
Deparmen of Economcs Unversy of Torono ECO408F M.A. Economercs Lecure Noes on Heeroskedascy Heeroskedascy o Ths lecure nvolves lookng a modfcaons we need o make o deal wh he regresson model when some of
More informationDynamic Team Decision Theory. EECS 558 Project Shrutivandana Sharma and David Shuman December 10, 2005
Dynamc Team Decson Theory EECS 558 Proec Shruvandana Sharma and Davd Shuman December 0, 005 Oulne Inroducon o Team Decson Theory Decomposon of he Dynamc Team Decson Problem Equvalence of Sac and Dynamc
More informationLet s treat the problem of the response of a system to an applied external force. Again,
Page 33 QUANTUM LNEAR RESPONSE FUNCTON Le s rea he problem of he response of a sysem o an appled exernal force. Agan, H() H f () A H + V () Exernal agen acng on nernal varable Hamlonan for equlbrum sysem
More informationMachine Learning Linear Regression
Machne Learnng Lnear Regresson Lesson 3 Lnear Regresson Bascs of Regresson Leas Squares esmaon Polynomal Regresson Bass funcons Regresson model Regularzed Regresson Sascal Regresson Mamum Lkelhood (ML)
More informationLecture 6: Learning for Control (Generalised Linear Regression)
Lecure 6: Learnng for Conrol (Generalsed Lnear Regresson) Conens: Lnear Mehods for Regresson Leas Squares, Gauss Markov heorem Recursve Leas Squares Lecure 6: RLSC - Prof. Sehu Vjayakumar Lnear Regresson
More informationHidden Markov Models Following a lecture by Andrew W. Moore Carnegie Mellon University
Hdden Markov Models Followng a lecure by Andrew W. Moore Carnege Mellon Unversy www.cs.cmu.edu/~awm/uorals A Markov Sysem Has N saes, called s, s 2.. s N s 2 There are dscree meseps, 0,, s s 3 N 3 0 Hdden
More informationVolatility Interpolation
Volaly Inerpolaon Prelmnary Verson March 00 Jesper Andreasen and Bran Huge Danse Mares, Copenhagen wan.daddy@danseban.com brno@danseban.com Elecronc copy avalable a: hp://ssrn.com/absrac=69497 Inro Local
More informationClustering (Bishop ch 9)
Cluserng (Bshop ch 9) Reference: Daa Mnng by Margare Dunham (a slde source) 1 Cluserng Cluserng s unsupervsed learnng, here are no class labels Wan o fnd groups of smlar nsances Ofen use a dsance measure
More informationMath 128b Project. Jude Yuen
Mah 8b Proec Jude Yuen . Inroducon Le { Z } be a sequence of observed ndependen vecor varables. If he elemens of Z have a on normal dsrbuon hen { Z } has a mean vecor Z and a varancecovarance marx z. Geomercally
More informationFiltrage particulaire et suivi multi-pistes Carine Hue Jean-Pierre Le Cadre and Patrick Pérez
Chaînes de Markov cachées e flrage parculare 2-22 anver 2002 Flrage parculare e suv mul-pses Carne Hue Jean-Perre Le Cadre and Parck Pérez Conex Applcaons: Sgnal processng: arge rackng bearngs-onl rackng
More informationOrdinary Differential Equations in Neuroscience with Matlab examples. Aim 1- Gain understanding of how to set up and solve ODE s
Ordnary Dfferenal Equaons n Neuroscence wh Malab eamples. Am - Gan undersandng of how o se up and solve ODE s Am Undersand how o se up an solve a smple eample of he Hebb rule n D Our goal a end of class
More informationAdvanced Machine Learning & Perception
Advanced Machne Learnng & Percepon Insrucor: Tony Jebara SVM Feaure & Kernel Selecon SVM Eensons Feaure Selecon (Flerng and Wrappng) SVM Feaure Selecon SVM Kernel Selecon SVM Eensons Classfcaon Feaure/Kernel
More informationTSS = SST + SSE An orthogonal partition of the total SS
ANOVA: Topc 4. Orhogonal conrass [ST&D p. 183] H 0 : µ 1 = µ =... = µ H 1 : The mean of a leas one reamen group s dfferen To es hs hypohess, a basc ANOVA allocaes he varaon among reamen means (SST) equally
More informationOn computing differential transform of nonlinear non-autonomous functions and its applications
On compung dfferenal ransform of nonlnear non-auonomous funcons and s applcaons Essam. R. El-Zahar, and Abdelhalm Ebad Deparmen of Mahemacs, Faculy of Scences and Humanes, Prnce Saam Bn Abdulazz Unversy,
More informationBayesian Inference of the GARCH model with Rational Errors
0 Inernaonal Conference on Economcs, Busness and Markeng Managemen IPEDR vol.9 (0) (0) IACSIT Press, Sngapore Bayesan Inference of he GARCH model wh Raonal Errors Tesuya Takash + and Tng Tng Chen Hroshma
More informationComb Filters. Comb Filters
The smple flers dscussed so far are characered eher by a sngle passband and/or a sngle sopband There are applcaons where flers wh mulple passbands and sopbands are requred Thecomb fler s an example of
More informationApproximate Analytic Solution of (2+1) - Dimensional Zakharov-Kuznetsov(Zk) Equations Using Homotopy
Arcle Inernaonal Journal of Modern Mahemacal Scences, 4, (): - Inernaonal Journal of Modern Mahemacal Scences Journal homepage: www.modernscenfcpress.com/journals/jmms.aspx ISSN: 66-86X Florda, USA Approxmae
More informationP R = P 0. The system is shown on the next figure:
TPG460 Reservor Smulaon 08 page of INTRODUCTION TO RESERVOIR SIMULATION Analycal and numercal soluons of smple one-dmensonal, one-phase flow equaons As an nroducon o reservor smulaon, we wll revew he smples
More informationRelative controllability of nonlinear systems with delays in control
Relave conrollably o nonlnear sysems wh delays n conrol Jerzy Klamka Insue o Conrol Engneerng, Slesan Techncal Unversy, 44- Glwce, Poland. phone/ax : 48 32 37227, {jklamka}@a.polsl.glwce.pl Keywor: Conrollably.
More informationCubic Bezier Homotopy Function for Solving Exponential Equations
Penerb Journal of Advanced Research n Compung and Applcaons ISSN (onlne: 46-97 Vol. 4, No.. Pages -8, 6 omoopy Funcon for Solvng Eponenal Equaons S. S. Raml *,,. Mohamad Nor,a, N. S. Saharzan,b and M.
More information( ) [ ] MAP Decision Rule
Announcemens Bayes Decson Theory wh Normal Dsrbuons HW0 due oday HW o be assgned soon Proec descrpon posed Bomercs CSE 90 Lecure 4 CSE90, Sprng 04 CSE90, Sprng 04 Key Probables 4 ω class label X feaure
More informationFI 3103 Quantum Physics
/9/4 FI 33 Quanum Physcs Aleander A. Iskandar Physcs of Magnesm and Phooncs Research Grou Insu Teknolog Bandung Basc Conces n Quanum Physcs Probably and Eecaon Value Hesenberg Uncerany Prncle Wave Funcon
More informationA NEW TECHNIQUE FOR SOLVING THE 1-D BURGERS EQUATION
S19 A NEW TECHNIQUE FOR SOLVING THE 1-D BURGERS EQUATION by Xaojun YANG a,b, Yugu YANG a*, Carlo CATTANI c, and Mngzheng ZHU b a Sae Key Laboraory for Geomechancs and Deep Underground Engneerng, Chna Unversy
More informationKernel-Based Bayesian Filtering for Object Tracking
Kernel-Based Bayesan Flerng for Objec Trackng Bohyung Han Yng Zhu Dorn Comancu Larry Davs Dep. of Compuer Scence Real-Tme Vson and Modelng Inegraed Daa and Sysems Unversy of Maryland Semens Corporae Research
More informationSampling Procedure of the Sum of two Binary Markov Process Realizations
Samplng Procedure of he Sum of wo Bnary Markov Process Realzaons YURY GORITSKIY Dep. of Mahemacal Modelng of Moscow Power Insue (Techncal Unversy), Moscow, RUSSIA, E-mal: gorsky@yandex.ru VLADIMIR KAZAKOV
More informationThis document is downloaded from DR-NTU, Nanyang Technological University Library, Singapore.
Ths documen s downloaded from DR-NTU, Nanyang Technologcal Unversy Lbrary, Sngapore. Tle A smplfed verb machng algorhm for word paron n vsual speech processng( Acceped verson ) Auhor(s) Foo, Say We; Yong,
More informationF-Tests and Analysis of Variance (ANOVA) in the Simple Linear Regression Model. 1. Introduction
ECOOMICS 35* -- OTE 9 ECO 35* -- OTE 9 F-Tess and Analyss of Varance (AOVA n he Smple Lnear Regresson Model Inroducon The smple lnear regresson model s gven by he followng populaon regresson equaon, or
More informationNew M-Estimator Objective Function. in Simultaneous Equations Model. (A Comparative Study)
Inernaonal Mahemacal Forum, Vol. 8, 3, no., 7 - HIKARI Ld, www.m-hkar.com hp://dx.do.org/.988/mf.3.3488 New M-Esmaor Objecve Funcon n Smulaneous Equaons Model (A Comparave Sudy) Ahmed H. Youssef Professor
More informationLecture 11 SVM cont
Lecure SVM con. 0 008 Wha we have done so far We have esalshed ha we wan o fnd a lnear decson oundary whose margn s he larges We know how o measure he margn of a lnear decson oundary Tha s: he mnmum geomerc
More information( t) Outline of program: BGC1: Survival and event history analysis Oslo, March-May Recapitulation. The additive regression model
BGC1: Survval and even hsory analyss Oslo, March-May 212 Monday May 7h and Tuesday May 8h The addve regresson model Ørnulf Borgan Deparmen of Mahemacs Unversy of Oslo Oulne of program: Recapulaon Counng
More informationLecture VI Regression
Lecure VI Regresson (Lnear Mehods for Regresson) Conens: Lnear Mehods for Regresson Leas Squares, Gauss Markov heorem Recursve Leas Squares Lecure VI: MLSC - Dr. Sehu Vjayakumar Lnear Regresson Model M
More informationPerformance Analysis for a Network having Standby Redundant Unit with Waiting in Repair
TECHNI Inernaonal Journal of Compung Scence Communcaon Technologes VOL.5 NO. July 22 (ISSN 974-3375 erformance nalyss for a Nework havng Sby edundan Un wh ang n epar Jendra Sngh 2 abns orwal 2 Deparmen
More informationRELATIONSHIP BETWEEN VOLATILITY AND TRADING VOLUME: THE CASE OF HSI STOCK RETURNS DATA
RELATIONSHIP BETWEEN VOLATILITY AND TRADING VOLUME: THE CASE OF HSI STOCK RETURNS DATA Mchaela Chocholaá Unversy of Economcs Braslava, Slovaka Inroducon (1) one of he characersc feaures of sock reurns
More informationNormal Random Variable and its discriminant functions
Noral Rando Varable and s dscrnan funcons Oulne Noral Rando Varable Properes Dscrnan funcons Why Noral Rando Varables? Analycally racable Works well when observaon coes for a corruped snle prooype 3 The
More information5th International Conference on Advanced Design and Manufacturing Engineering (ICADME 2015)
5h Inernaonal onference on Advanced Desgn and Manufacurng Engneerng (IADME 5 The Falure Rae Expermenal Sudy of Specal N Machne Tool hunshan He, a, *, La Pan,b and Bng Hu 3,c,,3 ollege of Mechancal and
More informationAttribute Reduction Algorithm Based on Discernibility Matrix with Algebraic Method GAO Jing1,a, Ma Hui1, Han Zhidong2,b
Inernaonal Indusral Informacs and Compuer Engneerng Conference (IIICEC 05) Arbue educon Algorhm Based on Dscernbly Marx wh Algebrac Mehod GAO Jng,a, Ma Hu, Han Zhdong,b Informaon School, Capal Unversy
More informationIntroduction ( Week 1-2) Course introduction A brief introduction to molecular biology A brief introduction to sequence comparison Part I: Algorithms
Course organzaon Inroducon Wee -2) Course nroducon A bref nroducon o molecular bology A bref nroducon o sequence comparson Par I: Algorhms for Sequence Analyss Wee 3-8) Chaper -3, Models and heores» Probably
More informationExistence and Uniqueness Results for Random Impulsive Integro-Differential Equation
Global Journal of Pure and Appled Mahemacs. ISSN 973-768 Volume 4, Number 6 (8), pp. 89-87 Research Inda Publcaons hp://www.rpublcaon.com Exsence and Unqueness Resuls for Random Impulsve Inegro-Dfferenal
More informationRobust and Accurate Cancer Classification with Gene Expression Profiling
Robus and Accurae Cancer Classfcaon wh Gene Expresson Proflng (Compuaonal ysems Bology, 2005) Auhor: Hafeng L, Keshu Zhang, ao Jang Oulne Background LDA (lnear dscrmnan analyss) and small sample sze problem
More informationFitting a Conditional Linear Gaussian Distribution
Fng a Condonal Lnear Gaussan Dsrbuon Kevn P. Murphy 28 Ocober 1998 Revsed 29 January 2003 1 Inroducon We consder he problem of fndng he maxmum lkelhood ML esmaes of he parameers of a condonal Gaussan varable
More informationGenetic Algorithm in Parameter Estimation of Nonlinear Dynamic Systems
Genec Algorhm n Parameer Esmaon of Nonlnear Dynamc Sysems E. Paeraks manos@egnaa.ee.auh.gr V. Perds perds@vergna.eng.auh.gr Ah. ehagas kehagas@egnaa.ee.auh.gr hp://skron.conrol.ee.auh.gr/kehagas/ndex.hm
More informationOnline Supplement for Dynamic Multi-Technology. Production-Inventory Problem with Emissions Trading
Onlne Supplemen for Dynamc Mul-Technology Producon-Invenory Problem wh Emssons Tradng by We Zhang Zhongsheng Hua Yu Xa and Baofeng Huo Proof of Lemma For any ( qr ) Θ s easy o verfy ha he lnear programmng
More informationUNIVERSITAT AUTÒNOMA DE BARCELONA MARCH 2017 EXAMINATION
INTERNATIONAL TRADE T. J. KEHOE UNIVERSITAT AUTÒNOMA DE BARCELONA MARCH 27 EXAMINATION Please answer wo of he hree quesons. You can consul class noes, workng papers, and arcles whle you are workng on he
More informationBayes rule for a classification problem INF Discriminant functions for the normal density. Euclidean distance. Mahalanobis distance
INF 43 3.. Repeon Anne Solberg (anne@f.uo.no Bayes rule for a classfcaon problem Suppose we have J, =,...J classes. s he class label for a pxel, and x s he observed feaure vecor. We can use Bayes rule
More informationLecture Slides for INTRODUCTION TO. Machine Learning. ETHEM ALPAYDIN The MIT Press,
Lecure Sldes for INTRDUCTIN T Machne Learnng ETHEM ALAYDIN The MIT ress, 2004 alpaydn@boun.edu.r hp://www.cmpe.boun.edu.r/~ehem/2ml CHATER 3: Hdden Marov Models Inroducon Modelng dependences n npu; no
More informationEcon107 Applied Econometrics Topic 5: Specification: Choosing Independent Variables (Studenmund, Chapter 6)
Econ7 Appled Economercs Topc 5: Specfcaon: Choosng Independen Varables (Sudenmund, Chaper 6 Specfcaon errors ha we wll deal wh: wrong ndependen varable; wrong funconal form. Ths lecure deals wh wrong ndependen
More informationNeural Networks-Based Time Series Prediction Using Long and Short Term Dependence in the Learning Process
Neural Neworks-Based Tme Seres Predcon Usng Long and Shor Term Dependence n he Learnng Process J. Puchea, D. Paño and B. Kuchen, Absrac In hs work a feedforward neural neworksbased nonlnear auoregresson
More informationHEAT CONDUCTION PROBLEM IN A TWO-LAYERED HOLLOW CYLINDER BY USING THE GREEN S FUNCTION METHOD
Journal of Appled Mahemacs and Compuaonal Mechancs 3, (), 45-5 HEAT CONDUCTION PROBLEM IN A TWO-LAYERED HOLLOW CYLINDER BY USING THE GREEN S FUNCTION METHOD Sansław Kukla, Urszula Sedlecka Insue of Mahemacs,
More informationM. Y. Adamu Mathematical Sciences Programme, AbubakarTafawaBalewa University, Bauchi, Nigeria
IOSR Journal of Mahemacs (IOSR-JM e-issn: 78-578, p-issn: 9-765X. Volume 0, Issue 4 Ver. IV (Jul-Aug. 04, PP 40-44 Mulple SolonSoluons for a (+-dmensonalhroa-sasuma shallow waer wave equaon UsngPanlevé-Bӓclund
More informationUse of Kalman Filtering and Particle Filtering in a Benzene Leachate Transport Model
Sudy of Cvl Engneerng and Archecure (SCEA Volume Issue 3, Sepember 03 www.sepub.org/scea Use of Kalman Flerng and Parcle Flerng n a Benzene Leachae Transpor Model Shoou-Yuh Chang, and Skdar Laf * Dep.
More informationJanuary Examinations 2012
Page of 5 EC79 January Examnaons No. of Pages: 5 No. of Quesons: 8 Subjec ECONOMICS (POSTGRADUATE) Tle of Paper EC79 QUANTITATIVE METHODS FOR BUSINESS AND FINANCE Tme Allowed Two Hours ( hours) Insrucons
More informationChapter Lagrangian Interpolation
Chaper 5.4 agrangan Inerpolaon Afer readng hs chaper you should be able o:. dere agrangan mehod of nerpolaon. sole problems usng agrangan mehod of nerpolaon and. use agrangan nerpolans o fnd deraes and
More informationHidden Markov Models
11-755 Machne Learnng for Sgnal Processng Hdden Markov Models Class 15. 12 Oc 2010 1 Admnsrva HW2 due Tuesday Is everyone on he projecs page? Where are your projec proposals? 2 Recap: Wha s an HMM Probablsc
More informationTime-interval analysis of β decay. V. Horvat and J. C. Hardy
Tme-nerval analyss of β decay V. Horva and J. C. Hardy Work on he even analyss of β decay [1] connued and resuled n he developmen of a novel mehod of bea-decay me-nerval analyss ha produces hghly accurae
More informationDEEP UNFOLDING FOR MULTICHANNEL SOURCE SEPARATION SUPPLEMENTARY MATERIAL
DEEP UNFOLDING FOR MULTICHANNEL SOURCE SEPARATION SUPPLEMENTARY MATERIAL Sco Wsdom, John Hershey 2, Jonahan Le Roux 2, and Shnj Waanabe 2 Deparmen o Elecrcal Engneerng, Unversy o Washngon, Seale, WA, USA
More informationIntroduction to Boosting
Inroducon o Boosng Cynha Rudn PACM, Prnceon Unversy Advsors Ingrd Daubeches and Rober Schapre Say you have a daabase of news arcles, +, +, -, -, +, +, -, -, +, +, -, -, +, +, -, + where arcles are labeled
More informationTHE PREDICTION OF COMPETITIVE ENVIRONMENT IN BUSINESS
THE PREICTION OF COMPETITIVE ENVIRONMENT IN BUSINESS INTROUCTION The wo dmensonal paral dfferenal equaons of second order can be used for he smulaon of compeve envronmen n busness The arcle presens he
More informationNotes on the stability of dynamic systems and the use of Eigen Values.
Noes on he sabl of dnamc ssems and he use of Egen Values. Source: Macro II course noes, Dr. Davd Bessler s Tme Seres course noes, zarads (999) Ineremporal Macroeconomcs chaper 4 & Techncal ppend, and Hamlon
More informationGraduate Macroeconomics 2 Problem set 5. - Solutions
Graduae Macroeconomcs 2 Problem se. - Soluons Queson 1 To answer hs queson we need he frms frs order condons and he equaon ha deermnes he number of frms n equlbrum. The frms frs order condons are: F K
More informationDensity Matrix Description of NMR BCMB/CHEM 8190
Densy Marx Descrpon of NMR BCMBCHEM 89 Operaors n Marx Noaon Alernae approach o second order specra: ask abou x magnezaon nsead of energes and ranson probables. If we say wh one bass se, properes vary
More informationHandout # 6 (MEEN 617) Numerical Integration to Find Time Response of SDOF mechanical system Y X (2) and write EOM (1) as two first-order Eqs.
Handou # 6 (MEEN 67) Numercal Inegraon o Fnd Tme Response of SDOF mechancal sysem Sae Space Mehod The EOM for a lnear sysem s M X DX K X F() () X X X X V wh nal condons, a 0 0 ; 0 Defne he followng varables,
More informationJ i-1 i. J i i+1. Numerical integration of the diffusion equation (I) Finite difference method. Spatial Discretization. Internal nodes.
umercal negraon of he dffuson equaon (I) Fne dfference mehod. Spaal screaon. Inernal nodes. R L V For hermal conducon le s dscree he spaal doman no small fne spans, =,,: Balance of parcles for an nernal
More informationAnalysis And Evaluation of Econometric Time Series Models: Dynamic Transfer Function Approach
1 Appeared n Proceedng of he 62 h Annual Sesson of he SLAAS (2006) pp 96. Analyss And Evaluaon of Economerc Tme Seres Models: Dynamc Transfer Funcon Approach T.M.J.A.COORAY Deparmen of Mahemacs Unversy
More informationSurvival Analysis and Reliability. A Note on the Mean Residual Life Function of a Parallel System
Communcaons n Sascs Theory and Mehods, 34: 475 484, 2005 Copyrgh Taylor & Francs, Inc. ISSN: 0361-0926 prn/1532-415x onlne DOI: 10.1081/STA-200047430 Survval Analyss and Relably A Noe on he Mean Resdual
More informationTraffic State Estimation from Aggregated Measurements with Signal Reconstruction Techniques
raffc Sae Esmaon from Aggregaed Measuremens wh Sgnal Reconsrucon echnques Vladmr Corc, Nemanja Djurc, and Slobodan Vucec he esmaon of he sae of raffc provdes a dealed pcure of he condons of a raffc nework
More informationA HIERARCHICAL KALMAN FILTER
A HIERARCHICAL KALMAN FILER Greg aylor aylor Fry Consulng Acuares Level 8, 3 Clarence Sree Sydney NSW Ausrala Professoral Assocae, Cenre for Acuaral Sudes Faculy of Economcs and Commerce Unversy of Melbourne
More informationON THE WEAK LIMITS OF SMOOTH MAPS FOR THE DIRICHLET ENERGY BETWEEN MANIFOLDS
ON THE WEA LIMITS OF SMOOTH MAPS FOR THE DIRICHLET ENERGY BETWEEN MANIFOLDS FENGBO HANG Absrac. We denfy all he weak sequenal lms of smooh maps n W (M N). In parcular, hs mples a necessary su cen opologcal
More information12d Model. Civil and Surveying Software. Drainage Analysis Module Detention/Retention Basins. Owen Thornton BE (Mech), 12d Model Programmer
d Model Cvl and Surveyng Soware Dranage Analyss Module Deenon/Reenon Basns Owen Thornon BE (Mech), d Model Programmer owen.hornon@d.com 4 January 007 Revsed: 04 Aprl 007 9 February 008 (8Cp) Ths documen
More informationDiscrete Markov Process. Introduction. Example: Balls and Urns. Stochastic Automaton. INTRODUCTION TO Machine Learning 3rd Edition
EHEM ALPAYDI he MI Press, 04 Lecure Sldes for IRODUCIO O Machne Learnng 3rd Edon alpaydn@boun.edu.r hp://www.cmpe.boun.edu.r/~ehem/ml3e Sldes from exboo resource page. Slghly eded and wh addonal examples
More informationComputer Robot Vision Conference 2010
School of Compuer Scence McGll Unversy Compuer Robo Vson Conference 2010 Ioanns Rekles Fundamenal Problems In Robocs How o Go From A o B? (Pah Plannng) Wha does he world looks lke? (mappng) sense from
More informationForecasting Using First-Order Difference of Time Series and Bagging of Competitive Associative Nets
Forecasng Usng Frs-Order Dfference of Tme Seres and Baggng of Compeve Assocave Nes Shuch Kurog, Ryohe Koyama, Shnya Tanaka, and Toshhsa Sanuk Absrac Ths arcle descrbes our mehod used for he 2007 Forecasng
More informationA Bayesian algorithm for tracking multiple moving objects in outdoor surveillance video
A Bayesan algorhm for racng mulple movng obecs n oudoor survellance vdeo Manunah Narayana Unversy of Kansas Lawrence, Kansas manu@u.edu Absrac Relable racng of mulple movng obecs n vdes an neresng challenge,
More informationImplementation of Quantized State Systems in MATLAB/Simulink
SNE T ECHNICAL N OTE Implemenaon of Quanzed Sae Sysems n MATLAB/Smulnk Parck Grabher, Mahas Rößler 2*, Bernhard Henzl 3 Ins. of Analyss and Scenfc Compung, Venna Unversy of Technology, Wedner Haupsraße
More informationTesting a new idea to solve the P = NP problem with mathematical induction
Tesng a new dea o solve he P = NP problem wh mahemacal nducon Bacground P and NP are wo classes (ses) of languages n Compuer Scence An open problem s wheher P = NP Ths paper ess a new dea o compare he
More informationCHAPTER 10: LINEAR DISCRIMINATION
CHAPER : LINEAR DISCRIMINAION Dscrmnan-based Classfcaon 3 In classfcaon h K classes (C,C,, C k ) We defned dscrmnan funcon g j (), j=,,,k hen gven an es eample, e chose (predced) s class label as C f g
More information. The geometric multiplicity is dim[ker( λi. number of linearly independent eigenvectors associated with this eigenvalue.
Lnear Algebra Lecure # Noes We connue wh he dscusson of egenvalues, egenvecors, and dagonalzably of marces We wan o know, n parcular wha condons wll assure ha a marx can be dagonalzed and wha he obsrucons
More informationLecture 18: The Laplace Transform (See Sections and 14.7 in Boas)
Lecure 8: The Lalace Transform (See Secons 88- and 47 n Boas) Recall ha our bg-cure goal s he analyss of he dfferenal equaon, ax bx cx F, where we emloy varous exansons for he drvng funcon F deendng on
More informationTools for Analysis of Accelerated Life and Degradation Test Data
Acceleraed Sress Tesng and Relably Tools for Analyss of Acceleraed Lfe and Degradaon Tes Daa Presened by: Reuel Smh Unversy of Maryland College Park smhrc@umd.edu Sepember-5-6 Sepember 28-30 206, Pensacola
More informationForecasting customer behaviour in a multi-service financial organisation: a profitability perspective
Forecasng cusomer behavour n a mul-servce fnancal organsaon: a profably perspecve A. Audzeyeva, Unversy of Leeds & Naonal Ausrala Group Europe, UK B. Summers, Unversy of Leeds, UK K.R. Schenk-Hoppé, Unversy
More information