Conservative Surrogate Model using Weighted Kriging Variance for Sampling-based RBDO

Size: px
Start display at page:

Download "Conservative Surrogate Model using Weighted Kriging Variance for Sampling-based RBDO"

Transcription

1 9 th World Congress on Structural and Multdsclnary Otmzaton June 13-17, 011, Shzuoka, Jaan Conservatve Surrogate Model usng Weghted Krgng Varance for Samlng-based RBDO Lang Zhao 1, K.K. Cho, Ikn Lee 3, Davd Lamb 4 and Davd Gorsch 5 1,,3 Deartment of Mechancal & Industral Engneerng College of Engneerng The Unversty of Iowa Iowa Cty, IA 54, U.S.A. Emal: lazhao@engneerng.uowa.edu kkcho@engneerng.uowa.edu lee@engneerng.uowa.edu 4,5 US Army RDECOM/TARDEC Warren, MI , U.S.A. davd.lamb@us.army.ml gorschd@tacom.army.ml 1. Abstract In samlng-based relablty-based desgn otmzaton of ractcal comlex engneerng alcatons, the Monte Carlo smulaton (MCS) for stochastc senstvty analyss and robablty of falure calculaton s based on the redcton from the surrogate model for the erformance functons. When the number of samles used to construct the surrogate model s small, the redcton from the surrogate model becomes naccurate and thus MCS becomes naccurate as well. Therefore, to count n the redcton error from the surrogate model and assure the obtaned otmum desgn from samlng-based RBDO does satsfy the robablstc constrants, a conservatve surrogate model s needed. In ths aer, a conservatve surrogate model s constructed usng the weghted Krgng varance where the weght s determned by the relatve change n the corrected Akake nformaton crteron (AICc) of the dynamc Krgng model. The roosed conservatve surrogate model erforms better than the tradtonal Krgng redcton nterval aroach because t does not generate unnecessary local otmum; and t erforms better than the constant safety margn aroach because t adatvely counts n the uncertanty of the surrogate model n the lace that the samles are sarse. Numercal examles show that usng the roosed conservatve surrogate model for samlng-based RBDO s necessary to assure the otmum desgn satsfes the robablstc constrants when the number of samles s lmted.. Keywords: conservatve surrogate model, dynamc Krgng method, weghted Krgng varance, corrected Akake nformaton crteron (AICc). 3. Introducton In samlng-based RBDO, Monte Carlo smulaton (MCS) s used to carry out both the robablty of falure and the stochastc senstvty analyss [1, ]. To carry out MCS effcently n comlex engneerng alcatons, a surrogate model s used to redct the erformance functon at the MCS samles. Zhao et al. [3] develoed a dynamc Krgng method to accurately construct the redcton of the erformance functon for the MCS by usng the genetc algorthm for the bass functon selecton and the generalzed attern search for the correlaton arameter estmaton. When the surrogate model s not accurate enough, new samles are sequentally nserted to mrove the redcton accuracy [3]. However, nsertng a large number of new samles sequentally may not be alcable when the samles are from the hyscal exerments or when the comutatonal resource for smulaton s lmted. Therefore, the mrovement of the surrogate model cannot be from the ncrease of the samle numbers. When alyng the surrogate model for relablty-based desgn otmzaton when the number of samles s small, to assure the obtaned otmum desgn can satsfy the robablstc constrants, a conservatve surrogate model s needed. Pcheny [4] used both safety margn and safety factor aroaches to construct the conservatve surrogate model and concluded that when the Krgng method s used, both methods rovde smlar erformance n terms of the conservatveness and the accuracy. Hertog et al. [5] and Luna and Young [6] used the bootstrang method to estmate the Krgng redcton nterval to construct the conservatve surrogate model. The bootstrang varance s larger than the tradtonal Krgng redcton varance by consderng the uncertanty from the correlaton arameter n the Krgng method. However, the bootstrang rocedure s tme-consumng and not alcable for hgh-dmensonal roblems. Vana et al. [7] used cross-valdaton to estmate the safety margn for the

2 conservatve surrogate model. Whle the cross-valdaton error s wdely used to estmate the redcton error, ths constant safety margn aroach does not dstngush the redcton error at dfferent samle locatons and wll yeld an over-conservatve surrogate model where the samles are aggregated around; and an under-conservatve surrogate model where the samles are sarse. In ths aer, a weghted Krgng varance s consdered to construct a more accurate conservatve surrogate model based the samle locatons. To evaluate the mortance from each samle and determne the weght for t, an accuracy measure of the surrogate model s needed frst. Under the Krgng framework, the Akake nformaton crteron (AIC) [8] s a relable ndcator to assess the accuracy of the surrogate model. Hurvch and Tsa [9] roosed a corrected AIC (AICc) to correct the bas n AIC when the number of samles s small. Burnham and Anderson [10] recommended usng AICc rather than AIC when the samle sze s small and showed that the AICc converges to AIC as the number of samles ncreases. Martn and Smson [11] comared the erformance of the corrected AIC assessment wth other methods and concluded that AICc rovdes the best accuracy assessment for surrogate model accuracy. In ths aer, the AICc s used to assess the accuracy changes n the surrogate model durng the cross-valdaton rocess, and an mortance functon value s assgned to each samle accordng to the relatve changes n AICc of the surrogate model and the weght s also assgned to each samle accordng to the mortance functon values. Then a weghted Krgng varance s calculated based on the weght values for each samle. By alyng ths weghted Krgng varance, one can construct the conservatve surrogate model for dynamc Krgng and use t for samlng-based RBDO to assure the obtaned otmum desgn can satsfy the robablstc constrant. The remander of ths aer s organzed n four sectons. rst, the backgrounds of the dynamc Krgng method, samlng-based RBDO, and the AICc are brefly summarzed. Then, the weghted Krgng varance usng the weght from the relatve change n AICc s ntroduced to construct the conservatve surrogate model. Thrd, the conservatve surrogate model s aled to samlng-based RBDO and the otmum desgn s verfed for the robablstc constrant usng numercal examles and comared wth the results usng the constant safety margn aroach. 4. Background of the Dynamc Krgng Method, Samlng-based RBDO, and AICc 4.1 Dynamc Krgng Method rst consder a unversal Krgng method. The outcomes are consdered as a realzaton of a stochastc rocess, and the redcted values are derved by alyng the stochastc rocess theory. Consder n samle onts T nd X [ x1, x,..., xn ] wth x R, and n resonses T Y [ y( x1), y( x),..., y( x n )] wth y( x ) R. In the Krgng method, the resonses at samles are consdered as a summaton of two arts as Y β e (1) The frst art of the rght-hand sde of Eq. (1), β, s consdered as the mean structure of the resonse, where =[ f( x ), 1,..., n] s a ( n K) f( x) fk ( x ), k 1,..., K reresents the user-defned bass functons, whch are usually n a smle olynomal form, such as1, xx,,.... The second art of the rght hand sde T n Eq. (1), e [ e( x1), e( x),..., e( x n )], s a realzaton of the stochastc rocess e( x ) that s assumed to have zero mean and covarance structure E[( e x)( e x)] R(, θ x, x ), where s the rocess varance and θ s the rocess correlaton arameter. The otmal choce of θ s defned as the maxmum lkelhood estmator (MLE), whch s the maxmzer of the lkelhood functon L, exressed as n 1 1 T 1 L R ex ( Yβ) R ( Yβ ) () where R s the symmetrc correlaton matrx wth - th comonent R R( θ, x, x ),, 1,..., n, and desgn matrx, and 1 ( ) T 1 ( ) T 1 Yβ R Yβ and 1 T 1 β R R Y are obtaned from the generalzed least square n regresson. Under the general decomoston of Eq. (1), the obectve s to redct the nose-free unbased resonse at a new ont of nterest x, exressed as T T 1 ( x) f β rr ( Yβ ) (3) where T 1 y krg r R( θ,x,x),..., R( θ,x,x). The redcton varance can be also obtaned as T 1 where u R rf. n ( ) (1 T ( T 1 ) 1 T 1 ) x u R ur R r (4)

3 In the dynamc Krgng method, the bass functon set f ( x) s no longer fxed. Instead, the otmal bass functon set s decded by the gven samles. rst, a number of olynomal functons are consdered as the canddate P functon. The hghest order P of the olynomal s decded by nd P n 1. A genetc algorthm s used to fnd the otmal subset of the bass functon by mnmzng the rocess varance. Second, a generalzed attern search algorthm s used to solve the roblem that s maxmzng Eq. (). Wth these two stes carred out, the dynamc Krgng can generate a more accurate surrogate model than the tradtonal unversal Krgng [3]. 4. Samlng-based RBDO The mathematcal formulaton of a general RBDO roblem s exressed as mnmze Cost( d) subect to Tar PG [ ( X) 0] P, 1,, nc (5) L U nd nr d dd, dr and XR T rv where d { d } μ(x ), 1~ nd s the desgn vector, whch s the mean value of the nd-dmensonal random rv T T varable vector X ={ X1, X,, X nd } ; ={ rv r rv r X X, X } where X and X stand for the random desgn Tar varable and random arameter comonents of the random nut X, resectvely; P s the target robablty of falure for the th constrant; and nc, nd, and nr are the number of robablstc constrants, desgn varables, and random varables lus arameters, resectvely. A relablty analyss for both the comonent and system levels nvolves calculaton of the robablty of falure, denoted by P, whch s defned usng a mult-dmensonal ntegral as P( ψ) P[ X] I ( ) f ( ; ) d EI ( ) nr x X x ψ x R X (6) where ψ s a vector of dstrbuton arameters, whch usually ncludes the mean (µ) and/or standard devaton (σ) of the random nut X X,, T 1 Xnr ; P reresents a robablty measure; s the falure set; f X ( x; ψ ) s a ont robablty densty functon (PD) of X; and E reresents the exectaton oerator. The falure set s defned as x: G ( ) 0 x for comonent relablty analyss of the th constrant functon G (x), and nc x: G ( ) 0 1 x and : nc ( ) 0 x G 1 x for the seres system and arallel system relablty analyss of nc erformance functons, resectvely. I ( x ) n Eq. (6) s called an ndcator functon and defned as 1, x I ( x ) (7) 0, otherwse In ths aer, snce the mean of X, μ,, T 1 nd s used as a desgn vector, the vector of dstrbuton arameters ψ s smly relaced wth µ for the comutaton of the robablty of falure n Eq. (6). Takng the artal dervatve of robablty of falure n Eq. (6) wth resect to the th desgn varable yelds P ( μ) I ( ) f ( ; ) d nr x R X x μ x (8) and the dfferental and ntegral oerators can be nterchanged usng the Lebnz s rule, gvng P ( μ) fx( x; μ) ln fx( x; μ) I ( ) ( ) ( ; ) nr d I f d nr x x x X x μ x R R (9) ln f ( ; ) X x μ EI ( x) snce I ( x ) s not a functon of. The artal dervatve of the log functon of the ont PD n Eq. (8) wth resect to s known as the frst-order score functon for and s denoted as (1) ln fx( ; ) s (; x μ ) x μ. (10) To comute the robablty of falure n Eq. (5) and the senstvty of robablty of falure n Eq. (8), statstcal samlng such as the Monte Carlo smulaton (MCS) at a gven desgn needs to be aled to true resonses, whch s comutatonally very exensve and almost rohbted. Hence, nstead of usng true resonses, whch are usually

4 obtaned from comuter smulatons, surrogate models are used be mlemented for the calculaton of the robablty of falure. To generate accurate surrogate models, ths aer uses the dynamc Krgng method, whch s dscussed n the revous secton. Denote the surrogate model obtaned by the dynamc Krgng method for the constrant functon G (X) as G ˆ ( X ). Then, by carryng out the MCS usng the conservatve surrogate model Gˆ ( X ), the robablstc constrants n Eq. (5) can be aroxmated as M 1 ( m) Tar P P[ G ( ) 0] ˆ ( ) X I x P (11) M ( m) where M s the MCS samle sze, x s the m th realzaton of X, and the falure set ˆ for the surrogate model s ˆ x: Gˆ ( x ) 0. Senstvty of the robablstc constrant n Eq. (8) s obtaned as defned as (1) ( m) where s ( x ; μ ) s obtaned usng Eq. (9). m1 P 1 x x μ (1) M ( m) (1) ( m) I ˆ ( ) s ( ; ) M m1 4.3 Corrected Akake nformaton crteron (AICc) AIC s orgnally roosed to evaluate the qualty of a model n statstcs based on the log-lkelhood functon [10]. It s a measure of the relatve goodness of ft of a statstcal model, and n general exressed as AIC k ln( L) (13) where k s the number of arameters n the statstcal model and L s the maxmzed value of the lkelhood functon for the estmated model. In the Krgng framework dscussed n Secton 4.1, the k s the number of dmenson of X, whch s nd, and L s the lkelhood functon as shown n Eq. (). When the samle sze s small,.e., n/k < 40 (n s the number of samle), whch s often the case n engneerng roblems, the corrected AIC (AICc) s used nstead to rovde an unbased estmaton, whch s exressed as kk ( 1) AICc AIC (14) n k 1 5 Samlng-based RBDO Usng Conservatve Surrogate Model 5.1 Weghted Krgng Predcton Varance for Conservatve Surrogate Model In Eq. (4), the Krgng redcton varance s calculated, and the C% uer bound of the redcton nterval s 1 C yvar ykrg (15) where s the nverse CD of the standard normal random varable. Ths redcton uer bound s usually used as the C% level conservatve surrogate model for the Krgng method, and often t s consdered as a varable safety margn aroach for constructng the conservatve surrogate model. Snce the Krgng method s an nterolaton method, the redcton varance becomes zero at the samle ont. As a result, the conservatve surrogate model y var s the same as the y krg at the samle onts. Ths nterolaton roerty of the Krgng method causes trouble when the conservatve surrogate model s used for an otmzaton roblem. Consder a 1-D examle, mn y (6x) sn(1x4) x [0,1] (16) The functon lot and the Krgng redcton usng seven onts are shown n g. 1.

5 wrong fgure!! gure 1: 1-D roblem wth 7 samles When the uer bound s used as the conservatve surrogate model and for otmzaton, the otmum x could be easly converged to the samle onts, whch are the local mnma of y var. Therefore, ths uer bound of the Krgng redcton nterval s not alcable for otmzaton. Another constant safety margn aroach s also often used for the conservatve surrogate model [7], where the safety margn s decded based on the emrcal CD of the cross-valdaton error. In ths constant safety margn aroach, the conservatve surrogate model s obtaned by shftng the Krgng redcton to a certan amount and s often exressed as yxv ykrg e XV,% C (17) where e XV,% C s the C ercentle of the cross-valdaton error. Ths conservatve surrogate model has the same dscreancy from the Krgng redcton regardless of the samle oston. Ths rases the roblem that the conservatve surrogate model cannot count n the dfferent uncertanty of the dscreancy due to dfferent samle locatons. Therefore, ths constant safety margn aroach may become over-conservatve f the samles are dense and under-conservatve f the samles are sarse. Consder the same examle shown n g.1. If there s no samle at x = 0.76, the safety margn Eq. (17) s s = and the conservatve surrogate model s shown n g.. The conservatve surrogate model s over-conservatve n the regon of [0, 0.55] and under-conservatve n the regon of [0.55, 1]. gure : 1-D roblem wth 6 samles

6 Accordng to the two examles dscussed above, one can see that when a conservatve surrogate model from the Krgng method s used for an otmzaton roblem, a varable safety margn aroach that has zero margn at the samle onts s not desrable because t generates unnecessary local otma regons; whereas a constant safety margn may not be desrable because t does not count n the effect from the samle oston. What s needed s a conservatve surrogate model that uses a varable safety margn that does not generate local otma and counts n the effect from the samle oston. In ths aer, a new conservatve surrogate model that combnes the varable safety factor aroach from Eq. (15) and the constant safety margn aroach from Eq. (17) s roosed. rst, to count n the effect from the samle oston, an mortance measure for each samle s needed. In ths aer, the AICc s chosen to quantfy how mortant a samle s for the Krgng redcton. The mortance functon for samle x s defned as ( ) AICc AICc Im( x ), 1,..., n (18) AICc ( ) where AICc s calculated usng all n samles and AICc s calculated by omttng x. rom Eq. (18), t shows that the larger Im( x ) s, the more mortant x s. Consder the same examle dscussed above. gs. 3(a)-(g) show the Krgng redctons when one of the samles s omtted, and the assocated mortance functon values are shown n g. 3(h) as well. The samle ont 6 at x = 0.76 s ndeed around the hghly nonlnear regon, and ts mortance functon value s the largest among all 7 samles. Therefore, ths mortance functon by the relatve change n AICc values can characterze how mortant one samle s for the Krgng redcton accuracy. (a) Krgng redcton w/o samle # (b) Krgng redcton w/o samle #3 (c) Krgng redcton w/o samle #4 (f) Krgng redcton w/o samle #5

7 (g) Krgng redcton w/o samle #6 (h) Im(x) functon values at each samle gure 3: Leave-one-out Krgng redcton After the mortance for each samle s calculated, to construct a weghted Krgng varance, a leave-one-out Krgng varance ( ) s calculated where the th samle x s omtted usng Eq. (4), and then the weghted Krgng varance s exressed as where the weght functon s defned as ( ) n weghted, wx ( ) 1 (19) wx ( ) 1/Im( x ) n 1 1/Im( x ) The weght functon decdes how much the leave-one-out Krgng varance ( ) contrbutes to the total Krgng varance when x s mssng accordng to the mortance functon value. nally, the conservatve surrogate model based on the weghted Krgng varance s 1 C ycon ykrg, weghted 100 (1) Accordng to Eq. (19), one can see that when x s omtted, the leave-one-out Krgng varance ( ) would not become zero at x, and therefore the total Krgng varance would not be zero at x and ndeed smoothes the conservatve surrogate model eventually. On the other hand, snce ( ) wll be larger at the lace where no samle s nearby, t wll make the total weghted Krgng varance become varable among the entre doman. Consder the same examle shown n gs. 4 and 5. In g. 4, the lnes labeled as Y_true, Y_krg, Y_xv,, Y_var and Y_con are the true resonse, the Krgng resonse, the conservatve surrogate model usng constant safety margn, the uer bound of Krgng redcton nterval, and the conservatve surrogate model usng the weghted Krgng varance, resectvely. When all seven samles are used and the conservatveness s set to be 90%, the conservatve surrogate model usng the weghted Krgng varance s smoother than the uer bound of the Krgng redcton nterval and closer to the true resonse than the conservatve surrogate model usng the safety margn. Moreover, f one mortant samle (.e., x = 0.76 n ths case) s mssng, the conservatve surrogate model usng the weghted Krgng varance would have a larger dscreancy than the conservatve surrogate model usng the constant safety margn aroach n the regon of [0.55, 1] and a smaller dscreancy n the regon of [0, 0.55], as shown n g. 5. These two cases ndcate that the conservatve surrogate model usng the weghted Krgng varance can adatvely dentfy the conservatveness accordng to samle locatons and rovde a smooth surrogate model that does not change the otmum regon of the orgnal resonse functon. (0)

8 gure 4: Conservatve surrogate model usng weghted Krgng varance (7 samles) gure 5: Conservatve surrogate model usng weghted Krgng varance (6 samles) 5.. Samlng-based RBDO usng the conservatve surrogate model When a lmted number of samles s used to generate the surrogate model for MCS n samlng-based RBDO, to assure the otmal desgn can satsfed the robablstc constrants, the surrogate models n Eqs. (11) and (1) need to be relaced by the conservatve surrogate model. Therefore, the conservatve surrogate model usng the weghted Krgng varance from Eq. (1) s used to reresent the orgnal erformance functon. The formulaton of samlng-based RBDO becomes mnmze Cost( d) ˆ c Tar subect to PG [ ( X) 0] P, 1,, nc () L U nd nr d dd dr XR, and where ˆ c G ( X) s the conservatve surrogate model for reresentng erformance functon G( X ). 6. Numercal Examle 6.1 -D RBDO Problem wth Hghly Nonlnear Constrant unctons Consder a -D RBDO roblem wth three robablstc constrants, exressed as

9 ( d d 10) ( d d 10) mnmze Cost( d) Tar subect to PG ( ( Xd ( )) 0) P.75%, 1 ~ 3 L U d d d, dr and XR where three constrants functons are X1 X G1 ( X) G ( X) 1 ( Y 6) ( Y 6) 0.6 ( Y 6) Z (4) 80 G3 ( X) 1 X1 8X 5 Y X1 where Z and are drawn n g. 6. The dstrbuton and desgn doman for each X varable are shown n Table 1. The ntal desgn ont s d 0 = [5 5]. Random Varables Table 1: Proertes of Random Varables Dstrbuton d L d 0 d U Standard Devaton X 1 Normal X Normal (3) gure 6: Cost and constrant functons lot It s worth mentonng that there s no need to construct the conservatve surrogate model from the ntal desgn ont. In the roosed RBDO rocess, the determnstc desgn otmzaton s carred out frst. The RBDO rocess starts from the determnstc otmum thereafter. The conservatve surrogate model wll be constructed for the RBDO rocess only. In ths examle, the determnstc otmum s x = [5.19, 0.74] (whch s the magenta cross n g. 7), and the number of samles n the local wndow s fxed to 10 samles from the Latn hyercube samlng method. Snce 10 samles may not be enough to construct the accurate surrogate models, the conservatve surrogate models usng the roosed weghted Krgng varance from Eq. (1) are generated for two actve constrants G 1 and G. or comarson study, the conservatve surrogate models usng the constant safety margn from Eq. (17) are also generated. The samlng-based RBDO s carred out usng these two dfferent conservatve surrogate models, and the otmum desgns are comared n Table and lotted n g. 7. The C% level s set to be 90% n ths examle. In g. 7, t can be seen that the surrogate model from the dynamc Krgng model tself, whch s the blue lne,

10 underestmates the true resonse and results n danger for the obtaned otmum desgn d = [4.743, ] (blue cross n g. 7) as the robablty of falure for G s.499%, whch s larger than the target robablty of falure.75%, as shown n Table. Therefore the conservatve surrogate model s ndeed necessary for countng n the uncertanty from the surrogate model and assurng that the otmum desgn can satsfy the robablstc constrants. By usng the weghted Krgng varance for the conservatve surrogate model, whch s the red lne n g.7, the obtaned RBDO otmum desgn s d = [4.700, ], whch s the red cross n g. 7, and the cost functon value, the robablty of falure for G 1, and the robablty of falure for G are ,.0654%, and %, resectvely. As a comarson, the constant safety margn aroach gves a more conservatve otmum desgn d = [4.6510, 1.605] (the green cross n g. 7) where the robablty of falure for G 1 and G are % and %, resectvely; and the cost functon value s rom the results n Table, one can see that the conservatve surrogate model usng the weghted Krgng varance can assure the otmum desgn satsfyng the robablstc constrants and has a better otmum desgn n terms of cost functon comared to the conservatve surrogate model usng the constant safety margn aroach. gure 7: Dfferent conservatve surrogate models and otmum desgns Table : Otmum desgns from dfferent surrogate models and the robablty of falure at the otmum Methods Cost Otmum Desgn P 1, % MCS (5M) P, % Dynamc Krgng , Surrogate Constant Safety Margn , Model Weghted Krgng Varance , Analytcal , When usng the surrogate model for samlng-based RBDO, the samle rofle may affect the result as well. A good samle rofle for dynamc Krgng may end u havng a better surrogate model. Therefore, to nvestgate whether the roosed weghted Krgng varance for the conservatve surrogate model has a stable erformance, a statstcal study s carred out. In ths statstcal study, 50 sets of 10-LHS samles are generated. or each samle set, the samlng-based RBDO s carred out usng both the weghted Krgng varance aroach and the constant safety margn aroach. The comarson for cost functon value and the robablty of falure at the otmum are shown n Table. 3. Methods Table: 3 Cost functon and robablty of falure at otmum desgn (50 trals) Cost (Medan) P 1, % (Medan) P, % (Medan) # of Volaton G 1 # of Volaton G Dynamc Krgng Constant Safety Margn

11 Weghted Krgng Varance In Table 3, when usng the dynamc Krgng method tself, the otmum desgns volate the robablstc constrant G 8 tmes out of 50 trals. It shows that the dynamc Krgng redcton ether underestmates the true resonse or overestmates the true resonse at a rough 50% chance for each way. By usng the weghted Krgng varance for the conservatve surrogate model, the obtaned otmum desgns volate the robablstc constrant G three tmes. As a comarson, f the constant safety margn s used for the conservatve surrogate model, the number of volatons for G s 1 tme out of 50 trals. Constrant G 1 s not volated n both conservatve surrogate modelng cases due to ts lnearty. The medan cost functon value at the otmum (whch s ) usng the constant safety margn s larger than the one by usng the weghted Krgng varance, whch s ( ). It s worth mentonng that snce the conservatveness level n ths examle s set to be 90%, both the constant safety margn aroach and the weghted Krgng varance aroach satsfy the conservatveness. However, the weghted Krgng varance aroach rovdes a better otmum n terms of the cost functon value. 7. Concluson When alyng the surrogate model for otmzaton roblems, the conservatve surrogate model usng the uer bound of the Krgng redcton nterval s not desrable because t generates multle local otmums at the samle onts. The conservatve surrogate model usng the constant safety margn does not dstngush the uncertanty of the surrogate model at dfferent samle locatons, and often the obtaned otmum becomes over-conservatve where samles are dense and under-conservatve where samles are sarse. A weghted Krgng varance usng the changes n AICc of the Krgng model to quantfy the mortance from each samle ont s roosed to construct a conservatve surrogate model that does not generate unnecessary local otmum and dstngushes the uncertanty of the surrogate model at dfferent samle locatons. By alyng the weghted Krgng varance for the conservatve surrogate model n the samlng-based RBDO roblem, the obtaned otmum satsfes the robablstc constrants at the conservatveness level and acheves a better otmum n terms of cost functon value comared wth the otmum usng the constant safety margn. 8. Acknowledgement Research s suorted by the Automotve Research Center, whch s sonsored by the U.S. Army Tank Automotve Research, Develoment and Engneerng Center (TARDEC) and Army Research Offce (ARO). 9. References [1] I. Lee., K.K. Cho., Y. Noh., L. Zhao., and D. Gorsch., Samlng-Based Stochastc Senstvty Analyss Usng Score unctons for RBDO Problems Wth Correlated Random Varables, Journal of Mechancal Desgn, , Vol. 133, 011. [] I. Lee., K.K. Cho., and L. Zhao., Samlng-Based RBDO Usng the Dynamc Krgng (D-Krgng) Method and Stochastc Senstvty Analyss, Submtted to Structural and Multdsclnary Otmzaton, 011. [3] L. Zhao., K.K., Cho., and I. Lee., A Metamodelng Method Usng Dynamc Krgng for Desgn Otmzaton, acceted by AIAA Journal, 011. [4] V. Pcheny., Imrovng Accuracy and Comensatng for Uncertanty n Surrogate Modelng, Ph.D Dssertaton, Unv. of lorda, Ganesvlle, L, 009. [5] D. Hertog., J. Klenen., and A. Sem., The correct Krgng varance estmated by bootstrang, Journal of the Oeratonal Research Socety, Vol. 57, , 006. [6] S.D. Luna and A. Young., The bootstra and Krgng Predcton Intervals, Scandnavan J Stat, Vol. 30, , 003. [7].A.C. Vana, V. Pcheny, and R.T. Haftka, "Usng cross valdaton to desgn conservatve surrogates," AIAA Journal, Vol. 48, No.10, , 010. [8] H. Akake., A new look at the statstcal model dentfcaton, IEEE Transacton on Automatc Control, Vol. 19, No. 6, , [9] C.M. Hurvch., and C.L. Tsa, Regresson and Tme Seres Model Selecton n Small Samles, Bometrka, Vol. 76, , [10] K.P. Burnham and D.R. Anderson., Model Selecton and Multmodel Inference: A Practcal Informaton-Theoretc Aroach, Srnger, 00. [11] J. Martn., and T.W. Smson., Use of Krgng Model to Aroxmate Determnstc Comuter Models, AIAA Journal, Vol. 43, No. 4, , 005.

Dr. Shalabh Department of Mathematics and Statistics Indian Institute of Technology Kanpur

Dr. Shalabh Department of Mathematics and Statistics Indian Institute of Technology Kanpur Analyss of Varance and Desgn of Exerments-I MODULE II LECTURE - GENERAL LINEAR HYPOTHESIS AND ANALYSIS OF VARIANCE Dr. Shalabh Deartment of Mathematcs and Statstcs Indan Insttute of Technology Kanur 3.

More information

A General Class of Selection Procedures and Modified Murthy Estimator

A General Class of Selection Procedures and Modified Murthy Estimator ISS 684-8403 Journal of Statstcs Volume 4, 007,. 3-9 A General Class of Selecton Procedures and Modfed Murthy Estmator Abdul Bast and Muhammad Qasar Shahbaz Abstract A new selecton rocedure for unequal

More information

Advanced Topics in Optimization. Piecewise Linear Approximation of a Nonlinear Function

Advanced Topics in Optimization. Piecewise Linear Approximation of a Nonlinear Function Advanced Tocs n Otmzaton Pecewse Lnear Aroxmaton of a Nonlnear Functon Otmzaton Methods: M8L Introducton and Objectves Introducton There exsts no general algorthm for nonlnear rogrammng due to ts rregular

More information

Dr. Shalabh Department of Mathematics and Statistics Indian Institute of Technology Kanpur

Dr. Shalabh Department of Mathematics and Statistics Indian Institute of Technology Kanpur Analyss of Varance and Desgn of Exerments-I MODULE III LECTURE - 2 EXPERIMENTAL DESIGN MODELS Dr. Shalabh Deartment of Mathematcs and Statstcs Indan Insttute of Technology Kanur 2 We consder the models

More information

Fuzzy approach to solve multi-objective capacitated transportation problem

Fuzzy approach to solve multi-objective capacitated transportation problem Internatonal Journal of Bonformatcs Research, ISSN: 0975 087, Volume, Issue, 00, -0-4 Fuzzy aroach to solve mult-objectve caactated transortaton roblem Lohgaonkar M. H. and Bajaj V. H.* * Deartment of

More information

Sampling-Based Stochastic Sensitivity Analysis Using Score Functions for RBDO Problems with Correlated Random Variables

Sampling-Based Stochastic Sensitivity Analysis Using Score Functions for RBDO Problems with Correlated Random Variables Proceedngs of the ASME 00 Internatonal Desgn Engneerng Techncal Conferences & Computers and Informaton n Engneerng Conference IDETC/CIE 00 August 5 8, 00, Montreal, Canada DETC00-859 Samplng-Based Stochastc

More information

Reliability-based design optimization using surrogate model with assessment of confidence level

Reliability-based design optimization using surrogate model with assessment of confidence level Unversty of Iowa Iowa Research Onlne Theses and Dssertatons Summer 2011 Relablty-based desgn optmzaton usng surrogate model wth assessment of confdence level Lang Zhao Unversty of Iowa Copyrght 2011 Lang

More information

Econ107 Applied Econometrics Topic 3: Classical Model (Studenmund, Chapter 4)

Econ107 Applied Econometrics Topic 3: Classical Model (Studenmund, Chapter 4) I. Classcal Assumptons Econ7 Appled Econometrcs Topc 3: Classcal Model (Studenmund, Chapter 4) We have defned OLS and studed some algebrac propertes of OLS. In ths topc we wll study statstcal propertes

More information

Managing Capacity Through Reward Programs. on-line companion page. Byung-Do Kim Seoul National University College of Business Administration

Managing Capacity Through Reward Programs. on-line companion page. Byung-Do Kim Seoul National University College of Business Administration Managng Caacty Through eward Programs on-lne comanon age Byung-Do Km Seoul Natonal Unversty College of Busness Admnstraton Mengze Sh Unversty of Toronto otman School of Management Toronto ON M5S E6 Canada

More information

A total variation approach

A total variation approach Denosng n dgtal radograhy: A total varaton aroach I. Froso M. Lucchese. A. Borghese htt://as-lab.ds.unm.t / 46 I. Froso, M. Lucchese,. A. Borghese Images are corruted by nose ) When measurement of some

More information

Dimension reduction method for reliability-based robust design optimization

Dimension reduction method for reliability-based robust design optimization Comuters and Structures xxx (2007) xxx xxx www.elsever.com/locate/comstruc Dmenson reducton method for relablty-based robust desgn otmzaton Ikjn Lee a, K.K. Cho a, *, Lu Du a, Davd Gorsch b a Deartment

More information

Bayesian Decision Theory

Bayesian Decision Theory No.4 Bayesan Decson Theory Hu Jang Deartment of Electrcal Engneerng and Comuter Scence Lassonde School of Engneerng York Unversty, Toronto, Canada Outlne attern Classfcaton roblems Bayesan Decson Theory

More information

Probabilistic Variation Mode and Effect Analysis: A Case Study of an Air Engine Component

Probabilistic Variation Mode and Effect Analysis: A Case Study of an Air Engine Component Probablstc Varaton Mode and Effect Analyss: A Case Study of an Ar Engne Comonent Pär Johannesson Fraunhofer-Chalmers Research Centre for Industral Mathematcs, Sweden; Par.Johannesson@fcc.chalmers.se Thomas

More information

Logistic regression with one predictor. STK4900/ Lecture 7. Program

Logistic regression with one predictor. STK4900/ Lecture 7. Program Logstc regresson wth one redctor STK49/99 - Lecture 7 Program. Logstc regresson wth one redctor 2. Maxmum lkelhood estmaton 3. Logstc regresson wth several redctors 4. Devance and lkelhood rato tests 5.

More information

Topology optimization of plate structures subject to initial excitations for minimum dynamic performance index

Topology optimization of plate structures subject to initial excitations for minimum dynamic performance index th World Congress on Structural and Multdsclnary Otmsaton 7 th -2 th, June 25, Sydney Australa oology otmzaton of late structures subject to ntal exctatons for mnmum dynamc erformance ndex Kun Yan, Gengdong

More information

Comparing two Quantiles: the Burr Type X and Weibull Cases

Comparing two Quantiles: the Burr Type X and Weibull Cases IOSR Journal of Mathematcs (IOSR-JM) e-issn: 78-578, -ISSN: 39-765X. Volume, Issue 5 Ver. VII (Se. - Oct.06), PP 8-40 www.osrjournals.org Comarng two Quantles: the Burr Tye X and Webull Cases Mohammed

More information

Confidence intervals for weighted polynomial calibrations

Confidence intervals for weighted polynomial calibrations Confdence ntervals for weghted olynomal calbratons Sergey Maltsev, Amersand Ltd., Moscow, Russa; ur Kalambet, Amersand Internatonal, Inc., Beachwood, OH e-mal: kalambet@amersand-ntl.com htt://www.chromandsec.com

More information

A Quadratic Cumulative Production Model for the Material Balance of Abnormally-Pressured Gas Reservoirs F.E. Gonzalez M.S.

A Quadratic Cumulative Production Model for the Material Balance of Abnormally-Pressured Gas Reservoirs F.E. Gonzalez M.S. Natural as Engneerng A Quadratc Cumulatve Producton Model for the Materal Balance of Abnormally-Pressured as Reservors F.E. onale M.S. Thess (2003) T.A. Blasngame, Texas A&M U. Deartment of Petroleum Engneerng

More information

Lecture Notes on Linear Regression

Lecture Notes on Linear Regression Lecture Notes on Lnear Regresson Feng L fl@sdueducn Shandong Unversty, Chna Lnear Regresson Problem In regresson problem, we am at predct a contnuous target value gven an nput feature vector We assume

More information

A Quadratic Cumulative Production Model for the Material Balance of Abnormally-Pressured Gas Reservoirs F.E. Gonzalez M.S.

A Quadratic Cumulative Production Model for the Material Balance of Abnormally-Pressured Gas Reservoirs F.E. Gonzalez M.S. Formaton Evaluaton and the Analyss of Reservor Performance A Quadratc Cumulatve Producton Model for the Materal Balance of Abnormally-Pressured as Reservors F.E. onale M.S. Thess (2003) T.A. Blasngame,

More information

Model Reference Adaptive Temperature Control of the Electromagnetic Oven Process in Manufacturing Process

Model Reference Adaptive Temperature Control of the Electromagnetic Oven Process in Manufacturing Process RECENT ADVANCES n SIGNAL PROCESSING, ROBOTICS and AUTOMATION Model Reference Adatve Temerature Control of the Electromagnetc Oven Process n Manufacturng Process JIRAPHON SRISERTPOL SUPOT PHUNGPHIMAI School

More information

Equivalent Standard Deviation to Convert High-reliability Model to Low-reliability Model for Efficiency of Samplingbased

Equivalent Standard Deviation to Convert High-reliability Model to Low-reliability Model for Efficiency of Samplingbased roceedngs of the ASME 0 Internatonal Desgn Engneerng echncal Conferences & Computers and Informaton n Engneerng Conference IDEC/CIE 0 August 8 3, 0, Washngton, D.C., USA DEC0-47537 Equvalent Standard Devaton

More information

ANSWERS. Problem 1. and the moment generating function (mgf) by. defined for any real t. Use this to show that E( U) var( U)

ANSWERS. Problem 1. and the moment generating function (mgf) by. defined for any real t. Use this to show that E( U) var( U) Econ 413 Exam 13 H ANSWERS Settet er nndelt 9 deloppgaver, A,B,C, som alle anbefales å telle lkt for å gøre det ltt lettere å stå. Svar er gtt . Unfortunately, there s a prntng error n the hnt of

More information

Complete Variance Decomposition Methods. Cédric J. Sallaberry

Complete Variance Decomposition Methods. Cédric J. Sallaberry Comlete Varance Decomoston Methods Cédrc J. allaberry enstvty Analyss y y [,,, ] [ y, y,, ] y ny s a vector o uncertan nuts s a vector o results s a comle uncton successon o derent codes, systems o de,

More information

Machine Learning. Classification. Theory of Classification and Nonparametric Classifier. Representing data: Hypothesis (classifier) Eric Xing

Machine Learning. Classification. Theory of Classification and Nonparametric Classifier. Representing data: Hypothesis (classifier) Eric Xing Machne Learnng 0-70/5 70/5-78, 78, Fall 008 Theory of Classfcaton and Nonarametrc Classfer Erc ng Lecture, Setember 0, 008 Readng: Cha.,5 CB and handouts Classfcaton Reresentng data: M K Hyothess classfer

More information

Response Surface Method Using Sequential Sampling for Reliability-Based Design Optimization

Response Surface Method Using Sequential Sampling for Reliability-Based Design Optimization Proceedngs of the ASME 9 Internatonal Desgn Engneerng echncal Conferences & Computers and Informaton n Engneerng Conference IDEC/CIE 9 August September, 9, San Dego, Calforna, USA DEC9-8784 Response Surface

More information

On New Selection Procedures for Unequal Probability Sampling

On New Selection Procedures for Unequal Probability Sampling Int. J. Oen Problems Comt. Math., Vol. 4, o. 1, March 011 ISS 1998-66; Coyrght ICSRS Publcaton, 011 www.-csrs.org On ew Selecton Procedures for Unequal Probablty Samlng Muhammad Qaser Shahbaz, Saman Shahbaz

More information

Linear Approximation with Regularization and Moving Least Squares

Linear Approximation with Regularization and Moving Least Squares Lnear Approxmaton wth Regularzaton and Movng Least Squares Igor Grešovn May 007 Revson 4.6 (Revson : March 004). 5 4 3 0.5 3 3.5 4 Contents: Lnear Fttng...4. Weghted Least Squares n Functon Approxmaton...

More information

Parametric fractional imputation for missing data analysis. Jae Kwang Kim Survey Working Group Seminar March 29, 2010

Parametric fractional imputation for missing data analysis. Jae Kwang Kim Survey Working Group Seminar March 29, 2010 Parametrc fractonal mputaton for mssng data analyss Jae Kwang Km Survey Workng Group Semnar March 29, 2010 1 Outlne Introducton Proposed method Fractonal mputaton Approxmaton Varance estmaton Multple mputaton

More information

Bayesian networks for scenario analysis of nuclear waste repositories

Bayesian networks for scenario analysis of nuclear waste repositories Bayesan networks for scenaro analyss of nuclear waste reostores Edoardo Toson ab Aht Salo a Enrco Zo bc a. Systems Analyss Laboratory Det of Mathematcs and Systems Analyss - Aalto Unversty b. Laboratory

More information

Algorithms for factoring

Algorithms for factoring CSA E0 235: Crytograhy Arl 9,2015 Instructor: Arta Patra Algorthms for factorng Submtted by: Jay Oza, Nranjan Sngh Introducton Factorsaton of large ntegers has been a wdely studed toc manly because of

More information

An Accurate Heave Signal Prediction Using Artificial Neural Network

An Accurate Heave Signal Prediction Using Artificial Neural Network Internatonal Journal of Multdsclnary and Current Research Research Artcle ISSN: 2321-3124 Avalale at: htt://jmcr.com Mohammed El-Dasty 1,2 1 Hydrograhc Surveyng Deartment, Faculty of Martme Studes, Kng

More information

SELECTION OF MIXED SAMPLING PLANS WITH CONDITIONAL DOUBLE SAMPLING PLAN AS ATTRIBUTE PLAN INDEXED THROUGH MAPD AND LQL USING IRPD

SELECTION OF MIXED SAMPLING PLANS WITH CONDITIONAL DOUBLE SAMPLING PLAN AS ATTRIBUTE PLAN INDEXED THROUGH MAPD AND LQL USING IRPD R. Samath Kumar, R. Vaya Kumar, R. Radhakrshnan /Internatonal Journal Of Comutatonal Engneerng Research / ISSN: 50 005 SELECTION OF MIXED SAMPLING PLANS WITH CONDITIONAL DOUBLE SAMPLING PLAN AS ATTRIBUTE

More information

REPORT DOCUMENTATION PAGE

REPORT DOCUMENTATION PAGE REPORT DOCUMENTATION PAGE orm Approved OMB NO. 0704-088 The publc reportng burden for ths collecton of nformaton s estmated to average hour per response, ncludng the tme for revewng nstructons, searchng

More information

Uncertainty as the Overlap of Alternate Conditional Distributions

Uncertainty as the Overlap of Alternate Conditional Distributions Uncertanty as the Overlap of Alternate Condtonal Dstrbutons Olena Babak and Clayton V. Deutsch Centre for Computatonal Geostatstcs Department of Cvl & Envronmental Engneerng Unversty of Alberta An mportant

More information

290 IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS I: FUNDAMENTAL THEORY AND APPLICATIONS, VOL. 45, NO. 3, MARCH H d (e j! ;e j!

290 IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS I: FUNDAMENTAL THEORY AND APPLICATIONS, VOL. 45, NO. 3, MARCH H d (e j! ;e j! 9 IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS I: FUNDAMENTAL THEORY AND APPLICATIONS, VOL. 45, NO. 3, MARCH 998 Transactons Brefs Two-Dmensonal FIR Notch Flter Desgn Usng Sngular Value Decomoston S.-C. Pe,

More information

Numerical studies of space filling designs: optimization algorithms and subprojection properties

Numerical studies of space filling designs: optimization algorithms and subprojection properties umercal studes of sace fllng desgns: otmzaton algorthms and subroecton roertes Bertrand Iooss wth Gullaume Dambln & Matheu Coulet CEMRACS 03 July, 30th, 03 Motvatng eamle: Uncertantes management n smulaton

More information

Statistical Inference. 2.3 Summary Statistics Measures of Center and Spread. parameters ( population characteristics )

Statistical Inference. 2.3 Summary Statistics Measures of Center and Spread. parameters ( population characteristics ) Ismor Fscher, 8//008 Stat 54 / -8.3 Summary Statstcs Measures of Center and Spread Dstrbuton of dscrete contnuous POPULATION Random Varable, numercal True center =??? True spread =???? parameters ( populaton

More information

( ) 2 ( ) ( ) Problem Set 4 Suggested Solutions. Problem 1

( ) 2 ( ) ( ) Problem Set 4 Suggested Solutions. Problem 1 Problem Set 4 Suggested Solutons Problem (A) The market demand functon s the soluton to the followng utlty-maxmzaton roblem (UMP): The Lagrangean: ( x, x, x ) = + max U x, x, x x x x st.. x + x + x y x,

More information

LINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity

LINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity LINEAR REGRESSION ANALYSIS MODULE IX Lecture - 30 Multcollnearty Dr. Shalabh Department of Mathematcs and Statstcs Indan Insttute of Technology Kanpur 2 Remedes for multcollnearty Varous technques have

More information

Probability, Statistics, and Reliability for Engineers and Scientists SIMULATION

Probability, Statistics, and Reliability for Engineers and Scientists SIMULATION CHATER robablty, Statstcs, and Relablty or Engneers and Scentsts Second Edton SIULATIO A. J. Clark School o Engneerng Department o Cvl and Envronmental Engneerng 7b robablty and Statstcs or Cvl Engneers

More information

Kernel Methods and SVMs Extension

Kernel Methods and SVMs Extension Kernel Methods and SVMs Extenson The purpose of ths document s to revew materal covered n Machne Learnng 1 Supervsed Learnng regardng support vector machnes (SVMs). Ths document also provdes a general

More information

An application of generalized Tsalli s-havrda-charvat entropy in coding theory through a generalization of Kraft inequality

An application of generalized Tsalli s-havrda-charvat entropy in coding theory through a generalization of Kraft inequality Internatonal Journal of Statstcs and Aled Mathematcs 206; (4): 0-05 ISS: 2456-452 Maths 206; (4): 0-05 206 Stats & Maths wwwmathsjournalcom Receved: 0-09-206 Acceted: 02-0-206 Maharsh Markendeshwar Unversty,

More information

Using Genetic Algorithms in System Identification

Using Genetic Algorithms in System Identification Usng Genetc Algorthms n System Identfcaton Ecaterna Vladu Deartment of Electrcal Engneerng and Informaton Technology, Unversty of Oradea, Unverstat, 410087 Oradea, Româna Phone: +40259408435, Fax: +40259408408,

More information

Combining Iterative Heuristic Optimization and Uncertainty Analysis methods for Robust Parameter Design

Combining Iterative Heuristic Optimization and Uncertainty Analysis methods for Robust Parameter Design Ths s a rernt of an artcle submtted for consderaton to the ournal Engneerng Otmzaton. It has been acceted and the revsed verson wll be avalable onlne at: htt://ournalsonlne.tandf.co.uk/ Combnng Iteratve

More information

The University of Iowa, Iowa City, IA 52242, USA. Warren, MI , USA. Dearborn, MI 48121, USA

The University of Iowa, Iowa City, IA 52242, USA. Warren, MI , USA. Dearborn, MI 48121, USA An Effcent Varable Screenng Method for Effectve Surrogate Models for Relablty-Based Desgn Optmzaton Hyunkyoo Cho 1, Sangjune Bae 1, K.K. Cho 1*, Davd Lamb, Ren-Jye Yang 3 1 Department of Mechancal and

More information

PID Controller Design Based on Second Order Model Approximation by Using Stability Boundary Locus Fitting

PID Controller Design Based on Second Order Model Approximation by Using Stability Boundary Locus Fitting PID Controller Desgn Based on Second Order Model Aroxmaton by Usng Stablty Boundary Locus Fttng Furkan Nur Denz, Bars Baykant Alagoz and Nusret Tan Inonu Unversty, Deartment of Electrcal and Electroncs

More information

Using T.O.M to Estimate Parameter of distributions that have not Single Exponential Family

Using T.O.M to Estimate Parameter of distributions that have not Single Exponential Family IOSR Journal of Mathematcs IOSR-JM) ISSN: 2278-5728. Volume 3, Issue 3 Sep-Oct. 202), PP 44-48 www.osrjournals.org Usng T.O.M to Estmate Parameter of dstrbutons that have not Sngle Exponental Famly Jubran

More information

Statistical analysis using matlab. HY 439 Presented by: George Fortetsanakis

Statistical analysis using matlab. HY 439 Presented by: George Fortetsanakis Statstcal analyss usng matlab HY 439 Presented by: George Fortetsanaks Roadmap Probablty dstrbutons Statstcal estmaton Fttng data to probablty dstrbutons Contnuous dstrbutons Contnuous random varable X

More information

A family of multivariate distributions with prefixed marginals

A family of multivariate distributions with prefixed marginals A famly of multvarate dstrbutons wth refxed margnals Isdro R. Cruz_Medna, F. Garca_Paez and J. R. Pablos_Tavares. Recursos Naturales, Insttuto Tecnológco de Sonora Cnco de Febrero 88, Cd. Obregón Son.

More information

Markov Chain Monte Carlo Lecture 6

Markov Chain Monte Carlo Lecture 6 where (x 1,..., x N ) X N, N s called the populaton sze, f(x) f (x) for at least one {1, 2,..., N}, and those dfferent from f(x) are called the tral dstrbutons n terms of mportance samplng. Dfferent ways

More information

Mixture of Gaussians Expectation Maximization (EM) Part 2

Mixture of Gaussians Expectation Maximization (EM) Part 2 Mture of Gaussans Eectaton Mamaton EM Part 2 Most of the sldes are due to Chrstoher Bsho BCS Summer School Eeter 2003. The rest of the sldes are based on lecture notes by A. Ng Lmtatons of K-means Hard

More information

Journal of Multivariate Analysis

Journal of Multivariate Analysis Journal of Multvarate Analyss 07 (202) 232 243 Contents lsts avalable at ScVerse ScenceDrect Journal of Multvarate Analyss journal homeage: www.elsever.com/locate/jmva James Sten tye estmators of varances

More information

STAT 3008 Applied Regression Analysis

STAT 3008 Applied Regression Analysis STAT 3008 Appled Regresson Analyss Tutoral : Smple Lnear Regresson LAI Chun He Department of Statstcs, The Chnese Unversty of Hong Kong 1 Model Assumpton To quantfy the relatonshp between two factors,

More information

Computing MLE Bias Empirically

Computing MLE Bias Empirically Computng MLE Bas Emprcally Kar Wa Lm Australan atonal Unversty January 3, 27 Abstract Ths note studes the bas arses from the MLE estmate of the rate parameter and the mean parameter of an exponental dstrbuton.

More information

On an Extension of Stochastic Approximation EM Algorithm for Incomplete Data Problems. Vahid Tadayon 1

On an Extension of Stochastic Approximation EM Algorithm for Incomplete Data Problems. Vahid Tadayon 1 On an Extenson of Stochastc Approxmaton EM Algorthm for Incomplete Data Problems Vahd Tadayon Abstract: The Stochastc Approxmaton EM (SAEM algorthm, a varant stochastc approxmaton of EM, s a versatle tool

More information

4DVAR, according to the name, is a four-dimensional variational method.

4DVAR, according to the name, is a four-dimensional variational method. 4D-Varatonal Data Assmlaton (4D-Var) 4DVAR, accordng to the name, s a four-dmensonal varatonal method. 4D-Var s actually a drect generalzaton of 3D-Var to handle observatons that are dstrbuted n tme. The

More information

Non-Ideality Through Fugacity and Activity

Non-Ideality Through Fugacity and Activity Non-Idealty Through Fugacty and Actvty S. Patel Deartment of Chemstry and Bochemstry, Unversty of Delaware, Newark, Delaware 19716, USA Corresondng author. E-mal: saatel@udel.edu 1 I. FUGACITY In ths dscusson,

More information

Digital PI Controller Equations

Digital PI Controller Equations Ver. 4, 9 th March 7 Dgtal PI Controller Equatons Probably the most common tye of controller n ndustral ower electroncs s the PI (Proortonal - Integral) controller. In feld orented motor control, PI controllers

More information

MACHINE APPLIED MACHINE LEARNING LEARNING. Gaussian Mixture Regression

MACHINE APPLIED MACHINE LEARNING LEARNING. Gaussian Mixture Regression 11 MACHINE APPLIED MACHINE LEARNING LEARNING MACHINE LEARNING Gaussan Mture Regresson 22 MACHINE APPLIED MACHINE LEARNING LEARNING Bref summary of last week s lecture 33 MACHINE APPLIED MACHINE LEARNING

More information

Multiple Regression Analysis

Multiple Regression Analysis Multle Regresson Analss Roland Szlág Ph.D. Assocate rofessor Correlaton descres the strength of a relatonsh, the degree to whch one varale s lnearl related to another Regresson shows us how to determne

More information

Power law and dimension of the maximum value for belief distribution with the max Deng entropy

Power law and dimension of the maximum value for belief distribution with the max Deng entropy Power law and dmenson of the maxmum value for belef dstrbuton wth the max Deng entropy Bngy Kang a, a College of Informaton Engneerng, Northwest A&F Unversty, Yanglng, Shaanx, 712100, Chna. Abstract Deng

More information

+, where 0 x N - n. k k

+, where 0 x N - n. k k CO 745, Mdterm Len Cabrera. A multle choce eam has questons, each of whch has ossble answers. A student nows the correct answer to n of these questons. For the remanng - n questons, he checs the answers

More information

Goodness of fit and Wilks theorem

Goodness of fit and Wilks theorem DRAFT 0.0 Glen Cowan 3 June, 2013 Goodness of ft and Wlks theorem Suppose we model data y wth a lkelhood L(µ) that depends on a set of N parameters µ = (µ 1,...,µ N ). Defne the statstc t µ ln L(µ) L(ˆµ),

More information

Lecture 2: Prelude to the big shrink

Lecture 2: Prelude to the big shrink Lecture 2: Prelude to the bg shrnk Last tme A slght detour wth vsualzaton tools (hey, t was the frst day... why not start out wth somethng pretty to look at?) Then, we consdered a smple 120a-style regresson

More information

Lesson 16: Basic Control Modes

Lesson 16: Basic Control Modes 0/8/05 Lesson 6: Basc Control Modes ET 438a Automatc Control Systems Technology lesson6et438a.tx Learnng Objectves Ater ths resentaton you wll be able to: Descrbe the common control modes used n analog

More information

Hidden Markov Model Cheat Sheet

Hidden Markov Model Cheat Sheet Hdden Markov Model Cheat Sheet (GIT ID: dc2f391536d67ed5847290d5250d4baae103487e) Ths document s a cheat sheet on Hdden Markov Models (HMMs). It resembles lecture notes, excet that t cuts to the chase

More information

2-Adic Complexity of a Sequence Obtained from a Periodic Binary Sequence by Either Inserting or Deleting k Symbols within One Period

2-Adic Complexity of a Sequence Obtained from a Periodic Binary Sequence by Either Inserting or Deleting k Symbols within One Period -Adc Comlexty of a Seuence Obtaned from a Perodc Bnary Seuence by Ether Insertng or Deletng Symbols wthn One Perod ZHAO Lu, WEN Qao-yan (State Key Laboratory of Networng and Swtchng echnology, Bejng Unversty

More information

Pattern Classification (II) 杜俊

Pattern Classification (II) 杜俊 attern lassfcaton II 杜俊 junu@ustc.eu.cn Revew roalty & Statstcs Bayes theorem Ranom varales: screte vs. contnuous roalty struton: DF an DF Statstcs: mean, varance, moment arameter estmaton: MLE Informaton

More information

Support Vector Machines. Vibhav Gogate The University of Texas at dallas

Support Vector Machines. Vibhav Gogate The University of Texas at dallas Support Vector Machnes Vbhav Gogate he Unversty of exas at dallas What We have Learned So Far? 1. Decson rees. Naïve Bayes 3. Lnear Regresson 4. Logstc Regresson 5. Perceptron 6. Neural networks 7. K-Nearest

More information

Possibility-Based Design Optimization Method for Design Problems with both Statistical and Fuzzy Input Data

Possibility-Based Design Optimization Method for Design Problems with both Statistical and Fuzzy Input Data 6 th World Congresses of Structural and Multdscplnary Optmzaton Ro de Janero, 30 May - 03 June 2005, Brazl Possblty-Based Desgn Optmzaton Method for Desgn Problems wth both Statstcal and Fuzzy Input Data

More information

On Unequal Probability Sampling Without Replacement Sample Size 2

On Unequal Probability Sampling Without Replacement Sample Size 2 Int J Oen Problems Com Math, Vol, o, March 009 On Unequal Probablt Samlng Wthout Relacement Samle Sze aser A Alodat Deartment of Mathematcs, Irbd atonal Unverst, Jordan e-mal: n_odat@ahoocom Communcated

More information

First Year Examination Department of Statistics, University of Florida

First Year Examination Department of Statistics, University of Florida Frst Year Examnaton Department of Statstcs, Unversty of Florda May 7, 010, 8:00 am - 1:00 noon Instructons: 1. You have four hours to answer questons n ths examnaton.. You must show your work to receve

More information

Global Sensitivity. Tuesday 20 th February, 2018

Global Sensitivity. Tuesday 20 th February, 2018 Global Senstvty Tuesday 2 th February, 28 ) Local Senstvty Most senstvty analyses [] are based on local estmates of senstvty, typcally by expandng the response n a Taylor seres about some specfc values

More information

Here is the rationale: If X and y have a strong positive relationship to one another, then ( x x) will tend to be positive when ( y y)

Here is the rationale: If X and y have a strong positive relationship to one another, then ( x x) will tend to be positive when ( y y) Secton 1.5 Correlaton In the prevous sectons, we looked at regresson and the value r was a measurement of how much of the varaton n y can be attrbuted to the lnear relatonshp between y and x. In ths secton,

More information

Matching Dyadic Distributions to Channels

Matching Dyadic Distributions to Channels Matchng Dyadc Dstrbutons to Channels G. Böcherer and R. Mathar Insttute for Theoretcal Informaton Technology RWTH Aachen Unversty, 5256 Aachen, Germany Emal: {boecherer,mathar}@t.rwth-aachen.de Abstract

More information

An Efficient Least-Squares Trilateration Algorithm for Mobile Robot Localization

An Efficient Least-Squares Trilateration Algorithm for Mobile Robot Localization he IEEE/RSJ Internatonal Conference on Intellgent Robots and Systems October -5, St. Lous, USA An Effcent Least-Squares rlateraton Algorthm for Moble Robot Localzaton Yu Zhou, Member, IEEE Abstract A novel

More information

Priority Queuing with Finite Buffer Size and Randomized Push-out Mechanism

Priority Queuing with Finite Buffer Size and Randomized Push-out Mechanism ICN 00 Prorty Queung wth Fnte Buffer Sze and Randomzed Push-out Mechansm Vladmr Zaborovsy, Oleg Zayats, Vladmr Muluha Polytechncal Unversty, Sant-Petersburg, Russa Arl 4, 00 Content I. Introducton II.

More information

Foundations of Arithmetic

Foundations of Arithmetic Foundatons of Arthmetc Notaton We shall denote the sum and product of numbers n the usual notaton as a 2 + a 2 + a 3 + + a = a, a 1 a 2 a 3 a = a The notaton a b means a dvdes b,.e. ac = b where c s an

More information

Basic Statistical Analysis and Yield Calculations

Basic Statistical Analysis and Yield Calculations October 17, 007 Basc Statstcal Analyss and Yeld Calculatons Dr. José Ernesto Rayas Sánchez 1 Outlne Sources of desgn-performance uncertanty Desgn and development processes Desgn for manufacturablty A general

More information

Statistics Chapter 4

Statistics Chapter 4 Statstcs Chapter 4 "There are three knds of les: les, damned les, and statstcs." Benjamn Dsrael, 1895 (Brtsh statesman) Gaussan Dstrbuton, 4-1 If a measurement s repeated many tmes a statstcal treatment

More information

4 Analysis of Variance (ANOVA) 5 ANOVA. 5.1 Introduction. 5.2 Fixed Effects ANOVA

4 Analysis of Variance (ANOVA) 5 ANOVA. 5.1 Introduction. 5.2 Fixed Effects ANOVA 4 Analyss of Varance (ANOVA) 5 ANOVA 51 Introducton ANOVA ANOVA s a way to estmate and test the means of multple populatons We wll start wth one-way ANOVA If the populatons ncluded n the study are selected

More information

Chapter 6. Supplemental Text Material

Chapter 6. Supplemental Text Material Chapter 6. Supplemental Text Materal S6-. actor Effect Estmates are Least Squares Estmates We have gven heurstc or ntutve explanatons of how the estmates of the factor effects are obtaned n the textboo.

More information

Composite Hypotheses testing

Composite Hypotheses testing Composte ypotheses testng In many hypothess testng problems there are many possble dstrbutons that can occur under each of the hypotheses. The output of the source s a set of parameters (ponts n a parameter

More information

Module 3 LOSSY IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur

Module 3 LOSSY IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur Module 3 LOSSY IMAGE COMPRESSION SYSTEMS Verson ECE IIT, Kharagpur Lesson 6 Theory of Quantzaton Verson ECE IIT, Kharagpur Instructonal Objectves At the end of ths lesson, the students should be able to:

More information

Classification as a Regression Problem

Classification as a Regression Problem Target varable y C C, C,, ; Classfcaton as a Regresson Problem { }, 3 L C K To treat classfcaton as a regresson problem we should transform the target y nto numercal values; The choce of numercal class

More information

Lecture 10 Support Vector Machines II

Lecture 10 Support Vector Machines II Lecture 10 Support Vector Machnes II 22 February 2016 Taylor B. Arnold Yale Statstcs STAT 365/665 1/28 Notes: Problem 3 s posted and due ths upcomng Frday There was an early bug n the fake-test data; fxed

More information

PHYS 450 Spring semester Lecture 02: Dealing with Experimental Uncertainties. Ron Reifenberger Birck Nanotechnology Center Purdue University

PHYS 450 Spring semester Lecture 02: Dealing with Experimental Uncertainties. Ron Reifenberger Birck Nanotechnology Center Purdue University PHYS 45 Sprng semester 7 Lecture : Dealng wth Expermental Uncertantes Ron Refenberger Brck anotechnology Center Purdue Unversty Lecture Introductory Comments Expermental errors (really expermental uncertantes)

More information

U-Pb Geochronology Practical: Background

U-Pb Geochronology Practical: Background U-Pb Geochronology Practcal: Background Basc Concepts: accuracy: measure of the dfference between an expermental measurement and the true value precson: measure of the reproducblty of the expermental result

More information

CHAPTER 5 NUMERICAL EVALUATION OF DYNAMIC RESPONSE

CHAPTER 5 NUMERICAL EVALUATION OF DYNAMIC RESPONSE CHAPTER 5 NUMERICAL EVALUATION OF DYNAMIC RESPONSE Analytcal soluton s usually not possble when exctaton vares arbtrarly wth tme or f the system s nonlnear. Such problems can be solved by numercal tmesteppng

More information

3.1 Expectation of Functions of Several Random Variables. )' be a k-dimensional discrete or continuous random vector, with joint PMF p (, E X E X1 E X

3.1 Expectation of Functions of Several Random Variables. )' be a k-dimensional discrete or continuous random vector, with joint PMF p (, E X E X1 E X Statstcs 1: Probablty Theory II 37 3 EPECTATION OF SEVERAL RANDOM VARIABLES As n Probablty Theory I, the nterest n most stuatons les not on the actual dstrbuton of a random vector, but rather on a number

More information

Segmentation Method of MRI Using Fuzzy Gaussian Basis Neural Network

Segmentation Method of MRI Using Fuzzy Gaussian Basis Neural Network Neural Informaton Processng - Letters and Revews Vol.8, No., August 005 LETTER Segmentaton Method of MRI Usng Fuzzy Gaussan Bass Neural Networ We Sun College of Electrcal and Informaton Engneerng, Hunan

More information

2E Pattern Recognition Solutions to Introduction to Pattern Recognition, Chapter 2: Bayesian pattern classification

2E Pattern Recognition Solutions to Introduction to Pattern Recognition, Chapter 2: Bayesian pattern classification E395 - Pattern Recognton Solutons to Introducton to Pattern Recognton, Chapter : Bayesan pattern classfcaton Preface Ths document s a soluton manual for selected exercses from Introducton to Pattern Recognton

More information

NUMERICAL DIFFERENTIATION

NUMERICAL DIFFERENTIATION NUMERICAL DIFFERENTIATION 1 Introducton Dfferentaton s a method to compute the rate at whch a dependent output y changes wth respect to the change n the ndependent nput x. Ths rate of change s called the

More information

Errors for Linear Systems

Errors for Linear Systems Errors for Lnear Systems When we solve a lnear system Ax b we often do not know A and b exactly, but have only approxmatons  and ˆb avalable. Then the best thng we can do s to solve ˆx ˆb exactly whch

More information

Estimation: Part 2. Chapter GREG estimation

Estimation: Part 2. Chapter GREG estimation Chapter 9 Estmaton: Part 2 9. GREG estmaton In Chapter 8, we have seen that the regresson estmator s an effcent estmator when there s a lnear relatonshp between y and x. In ths chapter, we generalzed the

More information

QUANTITATIVE RISK MANAGEMENT TECHNIQUES USING INTERVAL ANALYSIS, WITH APPLICATIONS TO FINANCE AND INSURANCE

QUANTITATIVE RISK MANAGEMENT TECHNIQUES USING INTERVAL ANALYSIS, WITH APPLICATIONS TO FINANCE AND INSURANCE QANTITATIVE RISK MANAGEMENT TECHNIQES SING INTERVA ANAYSIS WITH APPICATIONS TO FINANCE AND INSRANCE Slva DED Ph.D. Bucharest nversty of Economc Studes Deartment of Aled Mathematcs; Romanan Academy Insttute

More information

An Upper Bound on SINR Threshold for Call Admission Control in Multiple-Class CDMA Systems with Imperfect Power-Control

An Upper Bound on SINR Threshold for Call Admission Control in Multiple-Class CDMA Systems with Imperfect Power-Control An Upper Bound on SINR Threshold for Call Admsson Control n Multple-Class CDMA Systems wth Imperfect ower-control Mahmoud El-Sayes MacDonald, Dettwler and Assocates td. (MDA) Toronto, Canada melsayes@hotmal.com

More information

Dr. Shalabh Department of Mathematics and Statistics Indian Institute of Technology Kanpur

Dr. Shalabh Department of Mathematics and Statistics Indian Institute of Technology Kanpur Analyss of Varance and Desgn of Experment-I MODULE VII LECTURE - 3 ANALYSIS OF COVARIANCE Dr Shalabh Department of Mathematcs and Statstcs Indan Insttute of Technology Kanpur Any scentfc experment s performed

More information

Approximation of Optimal Interface Boundary Conditions for Two-Lagrange Multiplier FETI Method

Approximation of Optimal Interface Boundary Conditions for Two-Lagrange Multiplier FETI Method Aroxmaton of Otmal Interface Boundary Condtons for Two-Lagrange Multler FETI Method F.-X. Roux, F. Magoulès, L. Seres, Y. Boubendr ONERA, 29 av. de la Dvson Leclerc, BP72, 92322 Châtllon, France, ,

More information