A Comparative Theoretical and Computational Study on Robust Counterpart Optimization: III. Improving the Quality of Robust Solutions

Size: px
Start display at page:

Download "A Comparative Theoretical and Computational Study on Robust Counterpart Optimization: III. Improving the Quality of Robust Solutions"

Transcription

1 Ths s a ope access artcle publshed uder a ACS AuthorChoce Lcese, whch permts copyg ad redstrbuto of the artcle or ay adaptatos for o-commercal purposes. Artcle pubs.acs.org/iecr A Comparatve Theoretcal ad Computatoal Study o Robust Couterpart Optmzato: III. Improvg the Qualty of Robust Solutos Zuku L ad Chrstodoulos A. Floudas*, Departmet of Chemcal ad Materals Egeerg, Uversty of Alberta, Edmoto, AB T6G V4, Caada Departmet of Chemcal ad Bologcal Egeerg, Prceto Uversty, Prceto, New Jersey 08544, Uted States Dowloaded va o November 10, 018 at 15:4:38 (UTC). See for optos o how to legtmately share publshed artcles. ABSTRACT: I ths paper, we study the soluto qualty of robust optmzato problems whe they are used to approxmate probablstc costrats ad propose a ovel method to mprove the qualty. Two soluto frameworks are frst compared: (1) the tradtoal robust optmzato framework whch oly uses the a pror probablty bouds ad (3) the approxmato framework whch uses the a posteror probablty boud. We llustrate that the tradtoal robust optmzato method s computatoally effcet but ts soluto s geeral coservatve. O the other had, the a posteror probablty boud based method provdes less coservatve soluto but t s computatoally more dffcult because a ocovex optmzato problem s solved. Based o the comparatve study of the two methods, we propose a ovel teratve soluto framework whch combes the advatage of the a pror boud ad the a posteror probablty boud. The proposed method ca mprove the soluto qualty of tradtoal robust optmzato framework wthout sgfcatly creasg the computatoal effort. The effectveess of the proposed method s llustrated through umercal examples ad applcatos plag ad schedulg problems. 1. INTRODUCTION Data ucertaty wdely exsts realstc problems due to ther radom ature, measuremet errors, or other reasos. As a result, decso makg heretly volves cosderato of such ucertates sce the soluto of a optmzato problem ofte exhbts hgh sestvty to data perturbatos, ad gorg the ucertaty could lead to suboptmal or eve feasble solutos. I past decades, developg optmzato methods ad tools to facltate decso makg uder ucertaty has become oe of the most mportat topcs both the operatos research commuty ad also the process systems egeerg commuty. Robust optmzato belogs to a mportat methodology for dealg wth optmzato problems wth data ucertaty. Ths type of method eforces the costrat satsfacto for all possble realzatos of ucerta parameters sde a predefed ucertaty set. Comparg t to other methodologes that deal wth ucertaty, oe maor motvato of robust optmzato s that may applcatos the data set s a approprate oto of parameter ucertaty, especally for those cases that the parameter ucertaty s ot stochastc, or for staces where o dstrbutoal formato s avalable. Oe of the earlest papers o robust couterpart optmzato s the work of Soyster, 1 who cosdered smple perturbatos the data ad amed to fd a reformulato of the orgal lear programmg problem such that the resultg soluto would be feasble uder all possble perturbatos. The approach admts the hghest protecto ad s the most coservatve oe sce t esures feasblty agast all potetal realzatos. Thus, t s hghly desrable to provde a mechasm to allow for the trade-off betwee robustess ad performace. The work by Be-Tal ad Nemrovsk,,3 El-Ghaou et al., 4,5 ad Bertsmas ad Sm 6 vestgated the framework of robust couterpart optmzato by troducg dfferet types of ucertaty sets. Be-Tal ad Nemrovsk 3 proposed the ellpsodal set based robust couterpart formulato. El-Ghaou ad Lebret 4 troduced a robust optmzato approach for least-squares problems wth ucerta data. Bertsmas ad Sm 6 studed robust lear programmg wth coeffcet ucertaty usg a ucertaty set wth budgets. I ths robust couterpart optmzato formulato, a budget parameter (whch takes a value betwee zero ad the umber of ucerta coeffcet parameters the costrats ad s ot ecessarly teger) s troduced to cotrol the degree of coservatsm of the soluto. L et al. 7 ad Jaak et al. 8 developed the theory of the robust optmzato framework for geeral mxed-teger lear programmg problems ad cosdered both bouded ucertaty ad several kow probablty dstrbutos. The robust optmzato framework was later exteded by Verderame ad Floudas 9 ad they studed both cotuous (geeral, bouded, uform, ormal) ad dscrete (geeral, bomal, Posso) ucertaty dstrbutos ad appled the framework to operatoal plag problems. The work was further compared wth the codtoal value-at-rsk based method Verderame ad Floudas. 10 I the frst two parts of ths paper seres, 11,1 we systematcally studed the set duced robust couterpart optmzato techque for lear ad mxed teger lear optmzato problems. Dfferet ucertaty sets were extesvely studed, cludg those studed lterature ad ovel oes were troduced ths work. The relatoshp betwee dfferet represetatve ucertaty sets was dscussed, ad ther correspodg robust couterpart formulatos for both lear optmzato (LP) ad mxed teger lear optmzato (MILP) problems were derved. Probablstc guaratees Receved: May 8, 014 Revsed: Jue 18, 014 Accepted: July 3, 014 Publshed: July 3, Amerca Chemcal Socety 1311 dx.do.org/10.101/e Id. Eg. Chem. Res. 014, 53,

2 Idustral & Egeerg Chemstry Research o costrat satsfacto for the robust soluto of those dfferet ucertaty set duced robust couterpart optmzato models were derved, for both bouded ad ubouded ucertaty, wth ad wthout detal probablty dstrbuto formato. A key elemet applyg the robust optmzato framework s the selecto of the type ad the sze of the ucertaty set, whch s strogly related to the desred relablty of the soluto (.e., the probablty of costrat satsfacto). It s kow that chace costrat/probablstc costrat s the most drect way to eforce the relablty of the soluto of a optmzato problem, 13,18 where the relablty s expressed as a mmum requremet o the probablty of satsfyg costrats. Chace costraed optmzato problems face a lot of challeges for ther soluto. For example, eve evaluatg the dstrbuto of a sum of uformly dstrbuted depedet radom varables s very dffcult. 19 Whe the program has structural propertes that allow for a equvalet determstc formulato, a chace costraed problem ca be coverted to a determstc problem ad ca be solved drectly. 8 However, f the model does ot admt suffcet structure that ca be exploted, a approxmato method has to be used. The varous approxmato methods ca be dvded to samplg based methods ad aalytcal approxmato based methods. Frst, samplg based methods are desged based o the assumpto that t s possble to draw observatos from the dstrbuto of the ucertaty. Samplg based methods fall broadly to two categores: scearo approxmato ad sample average approxmato. For scearo approxmato, t draws a fte umber of samples from a gve dstrbuto, ad eforces all sampled costrats to hold. 14 Sample average approxmato refers to replacg the dstrbuto wth aother easy-to-use dstrbuto, typcally the emprcal dstrbuto determed from a sample draw from the orgal dstrbuto. 15 Whle solvg the approxmato problem represets oe aspect of complexty, the sze of the sample requred to guaratee the qualty of the approxmato s aother mportat lmtato. Secod, aalytcal approxmato methods are based o ether robust optmzato 16 or well-kow probablty equaltes. 17 Sce the type ad the sze of the ucertaty set s determed based o a tal assumpto o the costrat satsfacto ad the a pror probablty boud formulato, 1 robust optmzato provdes a safe approxmato of probablstc costrat. I cotrast to samplg based approxmato, robust optmzato based approxmato s a promsg determstc alteratve for certa classes of chace costraed problems. I addto, other forms of determstc aalytcal approxmato use probablty equaltes, such as the Markov equalty, Chebyshev s equalty, Berste s equalty, Hoeffdg s equalty, etc. Although robust optmzato has bee used wdely dfferet areas to acheve soluto robustess/relablty, the qualty of the soluto s ofte gored. I other words, whle the desred soluto feasblty (.e., desred probablty of costrat satsfacto) s met, how far s the soluto from optmalty? I ths work, we wll frst llustrate the above ssues ad the propose a teratve strategy for mprovg the robust soluto. I the proposed method, the tght a posteror probablty bouds are used to mprove the robust soluto wth a teratve framework. Compared to the sgle-step classcal robust optmzato method, the qualty of the robust soluto ca be mproved. O the other had, compared to the pure a posteror probablty boud based methods, the proposed method has the advatage that t does ot requre the global optmzato of ocovex problem. The rest of the paper s orgazed as follows. I secto, we frst preset the problem of optmzato wth probablstc guaratee o costrat satsfacto, that s, probablstcally costraed problem, ad the troduce the tradtoal robust optmzato based approxmato framework. Next, the a posteror probablty boud based approxmato framework s preseted secto 3. Both methods are studed through a umercal example. I secto 4, we preset a ovel teratve framework whch combes the advatage of the prevous two dfferet methods. The proposed method ad the tradtoal methods are studed through producto plag ad process schedulg problems secto 5, ad the paper s cocluded secto 6.. FRAMEWORK FOR ROBUST OPTIMIZATION.1. Problem Descrpto. Cosder the followg lear optmzato problem max cx s.t. ax + ax b J J (1) where x ad x ( = 1,..., ) ca be ether cotuous or teger varables, the left-had-sde (LHS) costrat coeffcets a ĩ are subect to ucertaty, ad J represets the dex subset that cotas the varable dces whose correspodg coeffcets are subect to ucertaty. The ucertates the costrat coeffcets are ormalzed by a ĩ = a + ξ a î J wth a beg the omal value ad a î beg a costat perturbato ampltude (a î > 0), {ξ } J are radom varables whch are subect to ucertaty. Wth the above defto, the th costrat problem 1 ca be rewrtte as the follows: ax + ξ ax b J Artcle () I may practcal applcatos, eforcg costrat satsfacto for all possble values of the ucerta parameters (.e., worst-case scearo) ca be too costly or eve mpossble. Probablstc costrat (also called chace costrat) provdes a compromse to avod ths stuato ad esures that the costrats are satsfed uder certa gve probablty. A probablstc verso of the above costrat s wrtte as follows so that a probablstc guaratee o costrat satsfacto s appled: Pr{ ax + ξax b} 1 ε J (3) or a upper boud o the probablty of costrat volato s appled Pr{ ax + ξax > b} ε J (4) where ε (0 < ε < 1) s the allowed degree of costrat volato. For stace, ε = 0.05 meas that the costrat must be satsfed wth a probablty larger tha 0.95 or the probablty of costrat volato must be less tha Whle ot probablstc costrats are alteratve for modelg soluto relablty, dvdual probablstc costrats are vestgated ths paper. Motvatg Example. Cosder the followg lear optmzato problem dx.do.org/10.101/e Id. Eg. Chem. Res. 014, 53,

3 Idustral & Egeerg Chemstry Research max 8x1 + 1x s.t. a11 x1 + a1 x 140 a1 x1 + a x 7 x1, x 0 ad assume that the LHS costrat coeffcets of the costrats are ucerta ad subect to ucertaty wth a 11 =10+ ξ 11, a 1 =0+ξ 1, a 1 =6+0.6ξ 1, ad a =8+0.8ξ ad ξ 11, ξ 1, ξ 1, ξ are depedet ucerta parameters uformly dstrbuted [ 1,1]. For the above problem, f we set the allowed volato probablty for each of the costrat as 0.05, the the probablstc costraed verso of the problem s max 8x + 1x 1 s.t. Pr{10x1 + 0 x + ( ξ11x1 + ξ1x) > 140} 0.05 Pr{6x1 + 8 x + (0.6ξ1x ξx) > 7} 0.05 x1, x 0.. Tradtoal Applcato Framework of Robust Optmzato. I set duced robust optmzato, the ucerta data s assumed to be varyg a gve ucertaty set ad the am s to choose the best soluto amog those mmuzed agast data ucertaty. For costrat, the set duced robust optmzato method ams to fd solutos that rema feasble for ay ξ the gve ucertaty set U so as to mmuze agast feasblty, that s ax + max ξ ax b ξ U J (5) The correspodg robust optmzato problem s max cx s.t. ax + max ξ ax b ξ U J (6) For dfferet ucertaty sets, the robust couterpart formulato s dstct. Furthermore, uder specfc probablty dstrbuto assumpto, the probablstc guaratee o the costrat satsfacto ca be quatfed usg the sze of the ucertaty set. I our prevous work, we have systematcally derved the robust couterpart formulatos uder dfferet ucertaty sets 11 ad also derved ther probablty bouds o costrat volato. 1 For example, f the ucertaty set s gve by a box U = { ξξ Ψ, J} (7) where Ψ s the sze of the box, the the robust optmzato couterpart costrat s ax +Ψ a x b J (8) If the ucerta parameters are subect to depedet bouded symmetrc dstrbuto, the the followg a pror probablty boud s vald Ψ = Ψ prorub prob ( ) exp volato (9) A pror probablty boud meas that f the sze of the box set s Ψ, the the soluto of the robust optmzato problem wll esure that the probablty of costrat volato s less tha or equal to the followg boud: Pr{ ax + ξ ax > b} prob ( Ψ) J prorub volato (10) I the lterature, the tradtoal way of applyg robust optmzato 7,4 to solve the probablstcally costraed problem s as follows. Frst, the relablty level ε the probablstc costrat s set, ad the type of the robust optmzato model (.e., ucertaty set) s selected by the dstrbuto of the ucertaty. Next, the sze of the ucertaty set s evaluated based o the a pror probablty bouds. For example, assumg that the box type ucertaty set s selected for applyg robust optmzato, the sze of the ucertaty set ca be determed by the followg problem m Ψ Ψ s.t.exp ε (11) Usg the sze parameter value determed from the above problem, the robust couterpart optmzato problem ca be solved ad the soluto esures that the costrat s satsfed wth the desred probablty 1 ε. max cx s.t. ax +Ψ a x b J Artcle (1) As a summary, the tradtoal framework of applyg robust optmzato to address the probablstc guaratee o costrat satsfacto s show Fgure 1. Fgure 1. Tradtoal framework of applyg robust optmzato 7,4 for probablstcally costraed optmzato problem. The robust optmzato problem provdes a safe ad coservatve approxmato of the probablstcally costraed problem. Notce that the mmum possble value for Ψ s used to fd the best possble soluto wth ths framework. I the followg, we llustrate the approxmato of the probablstc dx.do.org/10.101/e Id. Eg. Chem. Res. 014, 53,

4 Idustral & Egeerg Chemstry Research costraed problem usg the robust couterpart optmzato through the motvatg example. Motvatg Example (Cotued). The terval + ellpsodal ucertaty set duced robust couterpart optmzato model s appled to solve the motvatg example problem. The robust couterpart optmzato problem uder the terval + ellpsodal ucertaty set s as follows max 8x + 1x 1 s.t. 10x1 + 0x + u11 + u1 + Ω 1 z11 + 4z1 140 u x z u, u x z u x1 + 8x + 0.6u u +Ω 0.36z z 7 u x z u, u x z u x, x where Ω 1 ad Ω are parameters determg the sze of the terval + ellpsodal ucertaty set. Usg the probablty boud o costrat volato for ths type of robust couterpart optmzato model (uder the assumpto that the ucertaty s bouded ad symmetrc whch s satsfed ths example) Ω Ω exp 0.05, exp 0.05, B( J, Ω) 0.05 J we obtaed the smallest possble value: Ω 1 = Ω = The above robust optmzato problem s covex ad ca be effcetly solved usg covex olear optmzato solvers. Wth ths value, the robust couterpart optmzato problem ca be solved, the optmal obectve value s Ob* = ad the robust soluto s x = (7.77,.773). Ths soluto esures that the costrats are satsfed wth the desred probablty 0.95 o costrat satsfacto. Oce a robust soluto s obtaed, the probablty of costrat volato ca also be quatfed by a posteror probablty boud. I our prevous paper, 1 we studed those a posteror probablty bouds. If the probablstc dstrbuto formato o the ucerta parameters s kow, the the followg relatoshp holds: 1 + ξ > θ Pr{ ax ax b} exp( ( b ax) J θξ ax + l Ee [ ]) J (13) I the above dervato, θ s a arbtrary postve umber. As studed our prevous work, 1 wth the above probablty equalty, we ca evaluate the a posteror probablty boud o costrat volato as follows oce we have a set of soluto x, (.e., we have x as the soluto): Pr{ ax + ξax > b} m exp( θ( b a x ) + l E[ e ]) θ J J θξ ax (14) Notce that the above equato, a mmzato wth respect to θ (.e., oly oe varable) s performed to fd the tghtest probablty boud. For the tradtoal robust couterpart optmzato based framework, the adustable parameter defg the sze of the ucertaty set s tally selected based o the a pror probablty boud whch s a fucto of the adustable parameter. However, usually, the resultg soluto could be too coservatve, sce the actual probablty of costrat volato s much smaller tha the boud. For example, wth the robust soluto x = (7.77,.773) obtaed for the motvatg example, the probablty of costrat volato ca be evaluated usg the above a posteror probablty boud 14, ad the followg upper bouds o costrat volato for the two costrats ca be calculated: Pr{10x1 + 0 x + ( ξ11x1 + ξ1x ) > 140} Pr{6x1 + 8 x + (0.6ξ1x ξx ) > 7} whch are far less tha the desred volato probablty Ths mples that the obtaed robust soluto s coservatve ad there s room for mprovemet. 3. A POSTERIORI PROBABILITY BOUND BASED SOLUTION METHOD Whle the a posteror probablty boud ca be used to check the probablty of costrat satsfacto wth a gve soluto, t ca also be used aother way to formulate a safe approxmato of the probablstc costrat. Usg equalty 13, the followg safe approxmato of 4 s obtaed: θξ ax exp( θ( b a x) + l E[ e ]) ε J (15) because for ay feasble soluto satsfyg 15, t also satsfes the costrat 4. Costrat 15 ca be further rewrtte as θξ ax θ( b a x) + l E[ e ] l ε J (16) ad fally the followg safe approxmato of the probablstc costraed problem s obtaed: m cx x, θ θξ ax s.t. θ( b a x) + l E[ e ] l ε J Artcle (17) Note that whle for ay fxed value of θ > 0, 15 s a approxmato of the orgal probablstc costraed problem, here θ s set as a decso varables problem 17 so as to fd the tghtest possble approxmato ad to seek the best possble soluto. The a posteror probablty boud based framework addressg the probablstc costrat ca be represeted usg Fgure. It s see that the approxmato optmzato problem s costructed based o the selected a posteror boud, comparso to the selecto of ucertaty set the tradtoal robust optmzato framework. Defg ξ = ξ a î x, the the explct formulato of above approxmato problem depeds o the momet geeratg dx.do.org/10.101/e Id. Eg. Chem. Res. 014, 53,

5 Idustral & Egeerg Chemstry Research Fgure. Soluto framework of a posteror boud based method. Table 1. Summary o the pdf ad mgf of Some Dstrbutos dstrbuto of ξ uform U(a,b) fucto E[e θξ ]. For several kow dstrbutos, ther probablty desty fuctos ad momet geeratg fuctos are lsted Table 1, whch ca be substtuted to 17 ad the resultg problem ca be solved as a determstc optmzato problem. The above probablty equalty based approxmato framework s llustrated through the motvatg example. Motvatg Example (Cotued). Followg the dervato ths secto, the probablty equalty based safe approxmato of the probablstc costraed problem of the motvatg example s obtaed: max 8x + 1x x, θ 1 θξ x s.t. θ1(140 10x1 0 x) + l E[ e ] θξ 11x + l Ee [ ] l θξ x 1 1 θ(7 6x1 8 x) + l E[ e ] 0.8θξ x + l Ee [ ] l 0.05 x1, x 0, θ1, θ > 0 Sce the radom varable ξ 11 s subect to uform dstrbuto [ 1, 1], we have θξ e 111x1 Ee [ ] = probablty desty fucto f(ξ) θb θa 1/( b a), a ξ b e e 0, otherwse θ( b a) θ1x1 θ1x1 e θ x 1 1 momet geeratg fucto E(e θξ ) tragular θ θ ξ + 1, 1 ξ 0 e + e ξ + 1, 0 ξ 1 θ expoetal λξ 1 1 exp(λ) λe ξ 0 (1 θλ ) for λ θ 0 ξ < 0 ormal N(μ,σ) ξ 1 ( μ) θμ+ 0.5σ θ exp e πσ σ Evaluate the expectato terms the smlar way ad fally the followg problem s obtaed: max 8x + 1x x, θ 1 θx e e s.t. θ1(140 10x1 0 x) + l θ1x 1 θ1x θ1x e e + l l ε 4θ1x θ x e e θ(7 6x1 8 x) + l 1.θx1 0.8θx 0.8θx e e + l l ε 1.6θx x, x 0, θ, θ > θ1x θx1 The above problem s a ocovex optmzato problem, whch ca be solved through a determstc global optmzato approach. We solve the above problem through global optmzato solver ANTIGONE GAMS 4.. (wth relatve optmalty gap tolerace optcr = 0 ad resource lmt reslm = 10000) ad obta the followg soluto after s (wth a relatve gap 0.11% to the upper boud 9.33): Ob*= 9.9, x*= (7.3588,.7799), Artcle θ = (0.9501, 1.938) Comparg the above soluto wth the soluto from the tradtoal robust optmzato framework, t s observed that whle both solutos esure the desred probablty o costrat satsfacto, the a posteror probablty boud based method geerates a soluto whch s better tha the classcal method. Note though that the computatoal effort creases sce global optmzato s eeded. 4. ITERATIVE SOLUTION STRATEGY Comparg the prevous two methods, the followg observatos ca be made: (1) I terms of the formato eeded, the a posteror probablty boud based approxmato method eeds the exact probablty dstrbuto fucto whle the robust optmzato method oly eeds partal formato. For stace, the studed robust optmzato formulatos, the assumptos o ucertaty are oly bouded ad symmetrc so that a probablstc guaratee s vald. () I terms of the soluto complexty, the probablty equalty based approxmato problem ca be ocovex ad global optmzato s ecessary (.e., hgher computato complexty). The robust optmzato based approxmato leads to covex problem whch ca be solved very effcetly. (3) I terms of the qualty of the soluto, the a posteror probablty boud based approxmato method leads to less coservatve soluto because t s tghter tha the a pror probablty boud as llustrated prevous work. 1 The aforemetoed observatos show that there s a trade-off betwee the two dfferet types of approxmatos. To fully take advatage of both of them, a teratve soluto framework, whch s also the maor cotrbuto of ths paper, s proposed dx.do.org/10.101/e Id. Eg. Chem. Res. 014, 53,

6 Idustral & Egeerg Chemstry Research Artcle to combe the use of the tradtoal robust optmzato approxmato ad the a posteror probablty boud. The obectve s to mprove the qualty of robust soluto whle stll esure the probablstc guaratee of the robust soluto. At the same tme, the computatoal complexty s decreased comparg to the a posteror probablty boud based approxmato method. The proposed soluto framework s show Fgure 3, whch ca be detaled as follows. Itally, the type of the ucertaty set Notce that the proposed teratve framework, the problem of evaluatg the a posteror probablty boud usg the rghthad sde (RHS) of 14 s a optmzato problem. Sce ay feasble soluto of the RHS of 14 wll be a vald a posteror upper boud o the probablty of costrat volato, t s ot ecessary to obta the global optmal soluto here. Furthermore, sce x s a kow soluto ad take fxed value, the RHS of 14 s a sgle varable optmzato problem, whch ca be solved relatvely effcetly. Motvatg Example (Cotued). Applyg the proposed framework ad usg the terval + ellpsodal ucertaty set based robust couterpart optmzato formulato, we obta the followg results for the motvatg example as show Table for dfferet teratos (.e., k stads for terato): Fgure 3. Proposed teratve soluto framework to mprove the qualty of robust soluto. (.e., the robust optmzato formulato) s selected. Based o the desred degree of costrat satsfacto, the smallest possble sze of the ucertaty set s determed usg the specfc a pror probablty bouds (e.g.,exp( Ω /), exp( Ω / J ), etc.) of that type of robust formulato. Wth the determed sze of the ucertaty set, the robust couterpart optmzato problem s solved ad the robust soluto s obtaed. The, a upper boud o the costrat volato s evaluated usg the derved soluto ad the a posteror probablty boud. Ths probablty value s compared to the desred degree of costrat volato. If the gap betwee them s larger tha a certa predefed tolerace, the sze of the ucertaty set s adusted ad the robust optmzato problem s solved aga. The sze adustg s the mportat step the algorthm. Based o the fact that the a posteror probablty upper boud s mootocally decreasg fucto of the set sze, 1 ths adustmet ca be made heurstcally: f the probablty upper boud exceeds the desred level, the set sze should be decreased; f the boud s below the desred level, the set sze should be creased. The above procedure s repeated utl the gap betwee the desred degree of costrat volato ad the computed upper boud o costrat volato s less tha the tolerace. Fally, the soluto from the last roud robust optmzato step s reported. A pseudo code of the teratve algorthm s gve as follows: Table. Soluto Procedure k Ω 1,Ω exp( Ω posterorub /) Ob* (x 1, x ) prob volato 1 (.4477,.447) (1.38, 1.38) 3 (0.6119, ) 4 (0.9179, ) 5 (1.0709, ) 6 (1.1474, ) 7 (1.1856, ) (0.05, 0.05) (7.77,.773) (0.479, (7.745, 0.479).8009) (0.893, (7.6045, 0.893).9049) (0.656, (7.4, 0.656).859) (0.5636, 9.71 (7.335, ).836) (0.5177, 9.36 (7.93, ).84) (0.495, (7.354, ).777) ( , ) (0.0305, 0.005) (0.5486, 0.546) (0., 0.05) (0.105, 0.086) (0.06, 0.045) (0.045, 0.045) The detaled soluto procedure s explaed as follows: Step 1: Italze Ω 1 satsfy =.4477, Ω 1 volate =0,Ω satsfy =.4477, Ω volate = 0 usg the a pror boud. Set tolerace parameter δ = Step : Set Ω 1 = Ω 1 satsfy =.477, Ω = Ω satsfy =.477. Iterato 1 Step 3: Solve the robust optmzato problem ad obta soluto Ob* = 90.91, x 1 = 7.77, x =.773. Step 4: Compute the a posteror probablty boud from soluto P 1 = , P = Step 5: Sce the probablty volato upper boud s less tha 0.05 for both costrats, there s room to cotract the ucertaty set ad mprove the soluto. So decrease the sze of both ucertaty dx.do.org/10.101/e Id. Eg. Chem. Res. 014, 53,

7 Idustral & Egeerg Chemstry Research sets based o the value terato 1: Ω 1 = Ω = ( )/ = Iterato Step 3: Solve the robust model ad obta Ob* = , x 1 = 7.745, x = Step 4: Compute the a posteror probablty boud P 1 = , P = Step 5: Sce the probablty volato upper boud s stll less tha 0.05 for both costrats, t s ecessary to further decrease the sze of both ucertaty sets based o the sze value terato 1 Ω 1 = Ω = ( )/ = Iterato 3 Step 3: Solve the robust model ad get Ob* = , x 1 = , x = Step 4: Compute the a posteror probablty boud P 1 = , P = Step 5: Sce the probablty volato upper boud becomes larger tha 0.05 for both costrats, ths meas the ucertaty set should be elarged to satsfy chace costrat. So we adust the ucertaty sets based o smallest sze leadg to costrat satsfacto so far (1.38 terato ) ad the sze that leads to volato ( terato 3): Ω 1 = Ω = ( )/ = Iterato 4 Step 3: Solve the robust model ad obta Ob* = , x 1 = 7.4, x =.859. Step 4: Compute the a posteror probablty boud P 1 = 0., P = Step 5: Sce the probablty volato s stll larger tha 0.05 for both costrats by usg sze , we eed to further elarge the ucertaty set to satsfy chace costrat. We adust the parameter toward the smallest sze leadg to costrat satsfacto (1.38 terato ): Ω 1 = Ω = ( )/ = Iterato 5 Step 3: Solve the robust model ad obta Ob* = 9.71, x 1 = 7.335, x =.836. Step 4: Compute the a posteror probablty boud P 1 = 0.105, P = Step 5: Sce the probablty volato s stll larger tha 0.05 for both costrats the prevous terato, we eed to further elarge the ucertaty sets: Ω 1 = Ω = ( )/ = Iterato 6 Step 3: Solve the robust model ad obta Ob* = 9.36, x 1 = 7.93, x =.84. Step 4: Compute the a posteror probablty P 1 = 0.06, P = Step 5: Sce the probablty volato s larger tha 0.05 for the frst costrats ad the secod costrat s satsfed, we keep Ω uchaged ad further crease Ω 1 as Ω 1 = ( )/ = Iterato 7 Step 3: Solve the robust model ad obta Ob* = 9.153, x 1 = 7.354, x =.777. Step 4: Compute the a posteror probablty P 1 = 0.045, P = 0.045; both are less tha 0.05 ad the gap s smaller tha δ = 0.01, so the terato stops. Step 6: Retur the fal soluto Ob* = 9.153, x 1 = 7.354, x =.777. The soluto procedure of the above teratve method s also llustrated Fgure 4, whch shows how the costrat Fgure 4. Iteratve soluto procedure for the motvatg example: (upper) a posteror probablty boud; (lower) optmal obectve value of robust soluto. satsfacto probablty coverges to the desred level ad how the qualty of the robust soluto s evetually mproved comparg to the tradtoal framework. Fally, solutos from three dfferet methods for the motvatg example are summarzed Table 3. The colums Table 3. Comparg the Dfferet Solutos for the Motvatg Example Artcle tradtoal a posteror teratve Ob* (1.43%) (0.08%) posterorub prob volato ( , ) (0.05, 0.05) (0.045,0.045) CPU tme (s) tradtoal, a posteror, ad teratve represet the tradtoal robust optmzato framework, the a posteror probablty boud based method, ad the teratve method, respectvely. The row Ob* ad prob posterorub volato represet the optmal obectve value of the robust optmzato problem ad the a posteror probablty boud based o the robust soluto obtaed. The percetage umbers represet the gaps betwee the tradtoal/ teratve method s solutos ad the a posteror method s soluto. Comparg the tradtoal robust optmzato based approxmato framework ad the proposed teratve method, we observed that t mproves the qualty of the soluto whle stll esures the degree of costrat satsfacto. Notce that the robust soluto has bee mproved from to The percetage gap to the a posteror soluto 9.9 has bee decreased from 1.43% to 0.08% as show Table 3. Comparg the pure a posteror probablty boud based approxmato dx.do.org/10.101/e Id. Eg. Chem. Res. 014, 53,

8 Idustral & Egeerg Chemstry Research method ad the proposed teratve method, we observed that ts computatoal complexty s decreased, sce oly a set of covex robust optmzato problem s solved. The global optmzato of the ocovex optmzato s avoded. I addto, whe the gap tolerace s defed small eough, the soluto of the proposed method wll be close to the soluto of the probablty equalty based method. Fally, we summarze the characterstcs of the three methods Table 4. Table 4. Summary of Dfferet Methodologes tradtoal a posteror teratve ucertaty formato eeded partal full full soluto complexty low hgh low soluto qualty coservatve good good 5. CASE STUDIES I ths secto, we apply the dfferet methods to solve a producto plag problem ad a process schedulg problem to compare ther performaces ad llustrate the effectveess of the proposed teratve method. All the optmzato problems are solved o a UNIX workstato wth 3.40 GHz Itel Core CPU ad 8GB memory. The related global optmzato problems are solved va ANTIGONE ad the (mxed teger) lear optmzato problems are solved usg CPLEX 1.0 GAMS 4... Resource lmt s set as s for all cases Example 1. Ths example was troduced by L et al., 1 whch addresses the problem of plag the producto, storage ad marketg of a product for a compay. It s assumed that the compay eeds to make a producto pla for the comg year, dvded to sx perods of moths each, to maxmze the sales wth a gve cost budget. The producto cost cludes the cost of raw materal, labor, mache tme, etc., ad the cost fluctuates from perod to perod. The product maufactured durg a perod ca be sold the same perod, or stored ad sold later o. Operatos beg perod 1 wth a tal stock of 500 tos of the product storage, ad the compay would lke to ed up wth the same amout of the product storage at the ed of perod 6. Ths problem ca be formulated as a lear optmzato problem as follows: max Pz (18a) s.t. Cx + Vy (18b) x1 ( y + z ) = (18c) y + x ( y + z ) = 0 =,..., 6 1 (18d) y = (18e) x U = 1,..., 6 (18f) z D = 1,..., 6 x, y, z 0 = 1,..., 6 (18g) I ths example, t s assumed that the producto costs C are subect to depedet ucertaty dstrbutos. The ucertaty s ormalzed usg 50% of the omal value C as the base perturbato ampltude. The the orgal costrat 18b ca be rewrtte as ( C ξc) x + Vy where ξ are depedet radom varables. To esure the relablty of the soluto, the mmum probablty for costrat 18b to be satsfed s set as 0.85 (.e., the upper boud o the probablty of costrat volato s set to 0.15), the the probablstc costraed verso for ths costrat s Pr{ Cx + Vy > } 0.15 I the sequel, ths example s studed uder dfferet assumptos o the ucertaty dstrbutos. (a) Uform Dstrbuto. I ths case, t s assumed that the producto costs are subect to uform ucertaty, that s, ξ are radom varables that uformly dstrbutes [ 1, 1]. The tradtoal robust optmzato method s appled frst to solve the probablstcally costraed optmzato problem. Usg the terval + ellpsodal type ucertaty set, the followg robust couterpart optmzato costrats ca be formulated Cx + Vy + 0.5Cu + Ω 0.5C v u x v u The robust optmzato problem s obtaed by replacg the orgal determstc costrat 18b wth the above costrats. Usg the a pror probablty boud for the terval + ellpsodal set duced robust optmzato model, the sze of the ucertaty set s computed as Ω = The the robust optmzato problem s solved ad the correspodg obectve s The correspodg robust plag soluto s show Fgure 5. Fgure 5. Soluto of tradtoal robust optmzato method. The a posteror probablty boud based method s appled ext ad the followg costrat s appled to replace the orgal costrat 18b: 0.5θξCx θ( Cx Vy) + l E[ e ] l 0.15 For the a posteror probablty boud based method, the obectve value s It s see that ths soluto s better (hgher sales) tha that of the classcal robust optmzato method. The robust plag soluto s show Fgure 6. Artcle dx.do.org/10.101/e Id. Eg. Chem. Res. 014, 53,

9 Idustral & Egeerg Chemstry Research Fgure 6. Soluto of the a posteror probablty boud based approxmato model. Fally, the teratve framework s appled to solve the problem, the soluto procedure s show Table 5 ad the robust soluto s show Fgure 7. Table 5. Soluto Procedure of the Iteratve Method for Example 1 uder Uform Dstrbuto k Ω k exp( Ω posterorub /) Ob* prob volato Artcle reslm = s) s the best although global optmzato s ecessary. If we compare the classcal robust optmzato method ad the teratve robust optmzato method ad cosder the dfferece betwee ther soluto ad the best soluto (.e., the a posteror boud based soluto), the gap has bee decreased from to (a percetage of 9.93%), ad the soluto has bee greatly mproved through the teratve framework. Comparg the soluto of the a posteror probablty boud based method (Fgure 6) ad the teratve method (Fgure 7), t s see that the dfferece betwee the solutos s very small. Note that global optmzato s ot eeded the teratve framework ad oly fve covex robust couterpart optmzato problems are solved. (b) Tragular Dstrbuto. I ths case, t s assumed that the radom varables ξ are subect to symmetrc tragular dstrbuto wth support o [ 1,1]. Notce that ths type of dstrbuto s bouded ad symmetrc, so we ca stll apply the a pror probablty bouds to determe the sze of the ucertaty set ad the apply the tradtoal robust optmzato framework. Uder the terval + ellpsodal set duced robust optmzato model, the soluto wll be the same as the prevous uform dstrbuto sce the a pror probablty boud does ot deped o the dstrbuto. O the other had, the soluto of the other two methods wll chage sce they deped o the dstrbuto of the ucertaty. Specfcally, whle the a posteror probablty boud based method s appled, the soluto s after a s resource lmt reaches GAMS (wth a relatve gap of 7.5% to the upper boud ). Whe the teratve framework s appled, the fal soluto s ad the soluto procedure s show Table 7. Table 7. Soluto Procedure of Iteratve Method for Example 1 uder Tragular Dstrbuto Fgure 7. Soluto of teratve framework. I the above soluto procedure, the parameter Ω s adusted as follows: Ω = 0.5Ω 1, Ω 3 = 0.5(Ω 1 + Ω ), Ω 4 = 0.5(Ω + Ω 3 ). Notce that although the parameter value s decreased the fourth step, t s stll a coservatve soluto ( < 0.15), so we do ot adust the parameter usg Ω 5 = 0.5(Ω 3 + Ω 4 ), but rather Ω 5 = 0.5(Ω + Ω 4 ). The solutos of dfferet methods are summarzed the followg table. It s see from Table 6 that whle all the solutos Table 6. Results Summary of Example 1 uder Uform Dstrbuto tradtoal a posteror teratve Ob* (7.6%) (0.54%) posterorub prob volato CPU tme (s) satsfy the probablstc requremet o costrat satsfacto, the soluto of the a posteror probablty boud method (wth relatve optmalty gap tolerace optcr = 0 ad resource lmt k Ω k exp( Ω posterorub /) Ob* prob volato Table 8. Results Summary of Example 1 uder Tragular Dstrbuto tradtoal a posteror teratve Ob* (11.4%) (0.8%) posterorub prob volato CPU tme (s) The results of three dfferet methods are compared Table 8. Whle the tradtoal framework leads to a 11.4% dfferece to the a posteror method soluto, the teratve method s soluto oly has 0.8% dfferece. Ths shows that the teratve framework sgfcatly mprove the qualty of the robust soluto whle the relablty of the soluto s satsfed. I the above studes, t s assumed that the ucertaty dstrbuto s depedet, bouded ad symmetrc, such that we ca apply the tradtoal robust optmzato framework based o oly the a pror probablty bouds. However, whe the ucertaty dstrbuto does ot fall to ths characterstc, 1310 dx.do.org/10.101/e Id. Eg. Chem. Res. 014, 53,

10 Idustral & Egeerg Chemstry Research there s o bass for determg the sze of the ucertaty set. Cosequetly, the tradtoal robust optmzato framework caot be drectly appled (.e., f we stll use the a pror boud to determe the sze, the soluto wll ot esure the probablstc guaratee). O the other had, wth the proposed teratve framework, the robust optmzato ca stll be appled to solve the problem. Next, we study two cases where the dstrbuto does ot satsfy the bouded or symmetrc codto. (c) Expoetal Dstrbuto. I ths case, we assume the radom varables are subect to expoetal dstrbuto wth rate parameter λ = 1. Notce that ths dstrbuto s ubouded, so we apply the ellpsodal type ucertaty set duced robust optmzato model rather tha terval + ellpsodal type ths study. Wth the a posteror probablty boud based method, the fal soluto s Wth the teratve soluto framework, the soluto s ad the correspodg a posteror probablty upper boud of costrat volato s The soluto procedure s lsted Table 9. Notce that we do ot lst Table 9. Soluto Procedure of Iteratve Method for Example 1 uder Expoetal Dstrbuto posterorub k Ω k Ob* prob volato the a pror boud value here, sce t s ot applcable for the asymmetrc dstrbuto ths case. (d) Normal Dstrbuto. It s assumed that each ξ s subect to ormal dstrbuto N(0,0.5) ths case. Although for the case of the ormal dstrbuto, t s ot ecessary to apply a approxmato scheme to solve the probablstcally costraed problem sce aalytcal determstc equvalet problem ca be formulated ad solved, we study the robust optmzato approxmato based method here to compare the soluto qualty. The ellpsodal type ucertaty set s also used ths case to deal wth the ubouded dstrbuto. The a posteror method lead to soluto of , ad the teratve method leads to , wth the costrat volato probablty less tha as show Table 10. Table 10. Soluto Procedure of Iteratve Method for Example 1 uder Normal Dstrbuto posterorub k Ω k Ob* prob volato Example. I ths example, a process schedulg problem 11,0 s studed. Ths example volves the schedulg of a batch chemcal process related to the producto of two chemcal products usg three raw materals. The mxed teger lear optmzato model for the schedulg problem s formulated as follows, ad the readers are drected to the paper 11 for the detaled mxed teger lear optmzato formulato ad problem data max proft s.t. proft = prce d + prce (STI STF) s Sp, wv 1 I I,, s s, s Sr s s s C P ρ ρ st = st d b + b s S, (19a) (19b) s, s, 1 s, s,,, s,,, 1 Is J I J s st st s S, s, max s m max,,,,,,,, v wv b v wv I, J, d r s S s, s P Tf Ts + α wv + β b I, J,,,,,,,,,,, Ts Tf H(1 wv ),, + 1,,,, I, J, Ts Tf H(1 wv ),, + 1,,,,, I, J, Ts Tf H(1 wv ),, + 1,,,,, I,,, J, Ts Ts I, J, (19c) (19d) (19e) (19f) (19g) (19h) (19) (19),, + 1,, (19k) Tf Tf I, J,,, + 1,, (19l) Ts,, H I, J, (19m) Tf H I, J,,, (19) I ths example, we cosder the demad ucertaty oly ad the followg costrats are affected: r d 0 s S s s, P We assume depedet ucertaty dstrbutos o the product demad parameters r s ad assg a base perturbato of 0% of the omal demad data (r P1 = 50, r P = 75): r s = r s (1 + 0.ξ s ). We set the expected mmum probablty level o costrat satsfacto to 0.5 (.e., set the upper boud o costrat volato to 0.5). The the probablstc costraed verso s Pr{ r d > 0} 0.5 s S s s, Several dfferet type of ucertaty dstrbutos are cosdered to study the proposed method. P Artcle 1311 dx.do.org/10.101/e Id. Eg. Chem. Res. 014, 53,

11 Idustral & Egeerg Chemstry Research Artcle Fgure 8. Schedule obtaed from the a posteror boud based method. (a) Bouded Dstrbuto. We cosder a bouded symmetrc dstrbuto frst. Specfcally, we assume a uform dstrbuto o the product demad parameters r s (.e., ξ s s uformly dstrbuted [ 1,1]). Based o the tradtoal robust optmzato method, the robust couterpart optmzato costrat s formulated as follows: r d + Ψ(0. r) 0 s S s s, s s P whch s equvalet for box, ellpsodal, ad polyhedral type of ucertaty sets sce the umber of ucerta parameter each demad costrat s 1. From the desred boud o costrat volato 0.5, we get the ucertaty set sze value Sce the ucertaty s bouded, a set wth sze value 1 wll cover the whole ucertaty space. However, sze 1 stll makes the robust optmzato problem feasble. The a posteror probablty boud based method s appled ext ad the followg costrat s appled: 0. θξ s s r s θ( r d ) + l E[ e ] l 0.5 s S s s s, The resultg ocovex mxed teger olear optmzato problem s solved usg ANTIGONE (wth tolerace parameter optcr = 0.01 ad reslm = ) ad the obectve s (wth relatve optmalty gap 1.8% to the upper boud after s). The correspodg schedule s show Fgure 8. Fally, we apply the teratve soluto framework to solve the problem. Sce set sze value 1 makes the robust optmzato problem feasble, the teratve framework, we start from 0.5 for the parameter Ψ s to make the problem feasble. The soluto procedure s show Table 11. Notce that the Table 11. Soluto Procedure of Iteratve Method for Example uder Uform Dstrbuto k Ψ s exp( Ω posterorub /) Ob* prob volato 1 (1.0, 1.0) (0.6065, ) feasble (0.5, 0.5) (0.885, 0.885) (0.6646, 0.039) 3 (0.75, 0.5) (0.7548, 0.969) (0.3397, ) 4 (0.65, 0.375) (0.86, 0.931) (0.5070, 0.531) 5 (0.6875, 0.5) (0.7895, 0.969) (0.440, 0.839) 6 (0.6875, 0.15) (0.7895, 0.99) (0.440, 0.839) adustmets of the parameter values are based o the chage of the a posteror bouds. For example, we realze that for the frst demad costrat s the d terato ad the thrd terato. To move the boud close to less tha 0.5, we P set the ew parameter value Ψ P1 the fourth terato as ( )/ = 0.65, ad the resultg soluto lead to a ew probablty boud Notce the sxth step, there s o chage o the obectve soluto, so the soluto procedure s stopped. The fal optmal obectve value s , ad the correspodg schedule s show Fgure 9. Ths soluto has oly 0.54% dfferece to the a posteror soluto as show Table 1. Comparg the soluto from all the three dfferet methods show Table 1, the followg observatos ca be made. Frst, the tradtoal robust soluto framework s coservatve ad eve leads to a feasble problem. However, the teratve framework successfully addresses the same problem ad fds feasble soluto. The reaso s that the teratve framework utlzes ot oly the a pror probablty boud but also the a posteror probablty boud, thus avods the coservatve soluto. Secod, comparg the a posteror probablty boud based method ad the teratve method, t s see that the optmal solutos are very close. However, wth the teratve framework, we obta a soluto wth almost same qualty but far less computatoal efforts. Ths further valdates the effectveess of the proposed teratve method. (b) Ubouded Dstrbuto (Expoetal ad Normal). For ths schedulg example, two ubouded dstrbutos are also studed. The tradtoal framework s ot applcable ths stuato because the a pror probablty boud s based o the bouded dstrbuto assumpto. We frst study the expoetal dstrbuto wth parameter λ = 5. Box, ellpsodal, ad polyhedral type of ucertaty set lead to same robust optmzato formulatos here. The results of the dfferet methods are summarzed Table 13, ad the soluto procedure of teratve method s gve Table 14. The teratve method s soluto has oly 0.6% dfferece to the a posteror soluto ths case. Next, we study the case uder ormal dstrbuto N(0, 0.5). The results are gve Table 15, ad the soluto procedure of teratve method s summarzed Table 16. The a posteror soluto has a relatve gap of 1.33% to the upper boud after s. Iteratve method s soluto has oly 0.1% dfferece whe compared to the a posteror soluto. From the above results, t s observed that whle the tradtoal method s ot applcable to the ubouded dstrbuto cases, the teratve method apples the robust optmzato approxmato ad uses the a posteror boud to check the soluto relablty. The solutos of teratve method are cosstetly very close to the a posteror boud based method terms of the optmal obectve values whle the computatoal complexty s greatly reduced sce oly covex robust optmzato problem s solved several teratos. 131 dx.do.org/10.101/e Id. Eg. Chem. Res. 014, 53,

12 Idustral & Egeerg Chemstry Research Artcle Fgure 9. Schedule obtaed from the teratve method. Table 1. Results Summary of Example uder Uform Dstrbuto tradtoal a posteror teratve Ob* feasble (0.54%) posterorub prob volato (0.5, 0.558) (0.44, 0.839) CPU tme (s) Table 13. Results Summary of Example uder Expoetal Dstrbuto tradtoal a posteror teratve Ob* ot applcable (0.6%) posterorub prob volato (0.5, 0.167) (0.406, 0.187) CPU tme (s) Table 14. Soluto Procedure of Iteratve Method for Example uder Expoetal Dstrbuto posterorub k Ψ s Ob* prob volato 1 (1.0, 1.0) feasble (0.5, 0.5) (0.5578, ) 3 (0.75, 0.5) (0.397, 0.40) 4 (0.8, 0.) (0.1991, 0.394) 5 (0.7, 0.3) (0.873, 0.094) 6 (0.6, 0.4) (0.406, 0.187) Table 15. Results Summary of Example uder Normal Dstrbuto tradtoal a posteror teratve Ob* ot applcable (0.1%) posterorub prob volato (0.5, 0.547) (0.4868, 0.58) CPU tme (s) Table 16. Soluto Procedure of Iteratve Method for Example uder Normal Dstrbuto posterorub k Ψ s Ob* prob volato 1 (1.0, 1.0) feasble (0.5, 0.5) (0.6065, 0.357) 3 (0.75, 0.5) (0.347, 0.307) 4 (0.8, 0.) (0.780, 0.343) 5 (0.7, 0.3) (0.3753, 0.901) 6 (0.6, 0.4) (0.4868, 0.58) CONCLUSION The tradtoal robust optmzato framework ca be used to approxmate probablstc costrats ad provde safe soluto. However, the soluto ca be coservatve. Whe a detaled probablty dstrbuto o ucertaty s avalable, the a posteror probablty boud based method leads to less coservatve approxmato, but the trade-off s that the resultg ocovex problem eeds to be solved va a determstc global optmzato approach. A ovel soluto framework combg the robust optmzato approxmato ad the a posteror probablty boud evaluato s proposed to mprove the soluto qualty of tradtoal robust optmzato framework wthout sgfcat computato effort. The effectveess of the proposed method has bee llustrated through a motvatg example, as well as plag ad schedulg problems. Furthermore, whle the tradtoal robust optmzato method requres formato o certa probablty dstrbuto o the ucertaty such that the a pror probablty boud s vald, the proposed teratve framework exteds the applcato to geeral dstrbutos sce we ca always use the a posteror probablty boud to esure the costrat s satsfed wth desred probablty. Fally, t s worth metog that the probablty bouds used ths work are derved based o the assumpto of depedece o the ucerta parameters. Oe of the future research drectos wll be to vestgate the correlato betwee ucerta parameters. AUTHOR INFORMATION Correspodg Author *E-mal: floudas@tta.prceto.edu. Tel.: Fax: Notes The authors declare o competg facal terest. ACKNOWLEDGMENTS The authors gratefully ackowledge facal support from the Natoal Scece Foudato (CMMI ) ad the Natoal Isttute of Health (5R01LM009338). REFERENCES (1) Soyster, A. L. Covex programmg wth set-clusve costrats ad applcatos to exact lear programmg. Operatos Research 1973, 1, () Be-Tal, A.; Nemrovsk, A. Robust solutos of ucerta lear programs. Operatos Research Letters 1999, 5 (1), dx.do.org/10.101/e Id. Eg. Chem. Res. 014, 53,

13 Idustral & Egeerg Chemstry Research Artcle (3) Be-Tal, A.; Nemrovsk, A. Robust solutos of Lear Programmg problems cotamated wth ucerta data. Mathematcal Programmg 000, 88, (4) El-Ghaou, L.; Lebret, H. Robust solutos to least-squares problems wth ucerta data. Sam Joural o Matrx Aalyss ad Applcatos 1997, 18 (4), (5) El-Ghaou, L.; Oustry, F.; Lebret, H. Robust solutos to ucerta semdefte programs. SIAM J. Optm. 1998, 9, (6) Bertsmas, D.; Sm, M. The prce of robustess. Operatos Research 004, 5 (1), (7) L, X.; Jaak, S. L.; Floudas, C. A. A ew robust optmzato approach for schedulg uder ucertaty: I. bouded ucertaty. Comput. Chem. Eg. 004, 8, (8) Jaak, S. L.; L, X.; Floudas, C. A. A ew robust optmzato approach for schedulg uder ucertaty: II. Ucertaty wth kow probablty dstrbuto. Comput. Chem. Eg. 007, 31, (9) Verderame, P. M.; Floudas, C. A. Operatoal Plag of Large- Scale Idustral Batch Plats uder Demad Due Date ad Amout Ucertaty. I. Robust Optmzato Framework. Id. Eg. Chem. Res. 009, 48 (15), (10) Verderame, P. M.; Floudas, C. A. Operatoal Plag of Large- Scale Idustral Batch Plats uder Demad Due Date ad Amout Ucertaty: II. Codtoal Value-at-Rsk Framework. Id. Eg. Chem. Res. 010, 49 (1), (11) L, Z.; Dg, R.; Floudas, C. A. A Comparatve Theoretcal ad Computatoal Study o Robust Couterpart Optmzato: I. Robust Lear ad Robust Mxed Iteger Lear Optmzato. Id. Eg. Chem. Res. 011, 50, (1) L, Z.; Tag, Q.; Floudas, C. A. A Comparatve Theoretcal ad Computatoal Study o Robust Couterpart Optmzato: II. Probablstc Guaratees o Costrat Satsfacto. Id. Eg. Chem. Res. 01, 51, (13) Chares, A.; Cooper, W. W.; Symods, G. H. Cost Horzos ad Certaty Equvalets - a Approach to Stochastc-Programmg of Heatg Ol. Maagemet Scece 1958, 4 (3), (14) Nemrovsk, A.; Shapro, A., Scearo approxmatos of chace costrats. I Probablstc ad radomzed methods for desg uder ucertaty; Calafore, G.; Dabbee, F., Eds.; Sprger: Lodo, 005; pp (15) Luedtke, J.; Ahmed, S. A Sample Approxmato Approach for Optmzato wth Probablstc Costrats. SIAM Joural o Optmzato 008, 19 (), (16) Nemrovsk, A.; Shapro, A. Covex approxmatos of chace costraed programs. SIAM Joural o Optmzato 006, 17 (4), (17) Pter, J. Determstc approxmatos of probablty equaltes. ZOR - Method ad Models of Operatos Research 1989, 33, (18) Prekopa, A. Stochastc Programmg; Kluwer: Dordrecht, (19) Mtra, S. K. Probablty Dstrbuto of Sum of Uformly Dstrbuted Radom Varables. Sam Joural o Appled Mathematcs 1971, 0 (), (0) Ierapetrtou, M. G.; Floudas, C. A. Effectve cotuous-tme formulato for short-term schedulg.. Cotuous ad semcotuous processes. Id. Eg. Chem. Res. 1998, 37 (11), (1) Ierapetrtou, M. G.; Floudas, C. A. Short-term schedulg: New mathematcal models vs algorthmc mprovemets. Comput. Chem. Eg. 1998,, S419 S46. () Ierapetrtou, M. G.; Hee, T. S.; Floudas, C. A. Effectve cotuous-tme formulato for short-term schedulg. 3. Multple termedate due dates. Id. Eg. Chem. Res. 1999, 38 (9), (3) Mseer, R.; Floudas, C. A. ANTIGONE: Algorthms for contuous/iteger Global Optmzato of Nolear Equatos. Joural of Global Optmzato 014, 59, , DOI: / s (4) L, Z.; Ierapetrtou, M. G. Robust optmzato for process schedulg uder ucertaty. Id. Eg. Chem. Res. 008, 47, dx.do.org/10.101/e Id. Eg. Chem. Res. 014, 53,

Functions of Random Variables

Functions of Random Variables Fuctos of Radom Varables Chapter Fve Fuctos of Radom Varables 5. Itroducto A geeral egeerg aalyss model s show Fg. 5.. The model output (respose) cotas the performaces of a system or product, such as weght,

More information

Solving Constrained Flow-Shop Scheduling. Problems with Three Machines

Solving Constrained Flow-Shop Scheduling. Problems with Three Machines It J Cotemp Math Sceces, Vol 5, 2010, o 19, 921-929 Solvg Costraed Flow-Shop Schedulg Problems wth Three Maches P Pada ad P Rajedra Departmet of Mathematcs, School of Advaced Sceces, VIT Uversty, Vellore-632

More information

Summary of the lecture in Biostatistics

Summary of the lecture in Biostatistics Summary of the lecture Bostatstcs Probablty Desty Fucto For a cotuos radom varable, a probablty desty fucto s a fucto such that: 0 dx a b) b a dx A probablty desty fucto provdes a smple descrpto of the

More information

UNIVERSITY OF OSLO DEPARTMENT OF ECONOMICS

UNIVERSITY OF OSLO DEPARTMENT OF ECONOMICS UNIVERSITY OF OSLO DEPARTMENT OF ECONOMICS Postpoed exam: ECON430 Statstcs Date of exam: Jauary 0, 0 Tme for exam: 09:00 a.m. :00 oo The problem set covers 5 pages Resources allowed: All wrtte ad prted

More information

Point Estimation: definition of estimators

Point Estimation: definition of estimators Pot Estmato: defto of estmators Pot estmator: ay fucto W (X,..., X ) of a data sample. The exercse of pot estmato s to use partcular fuctos of the data order to estmate certa ukow populato parameters.

More information

Chapter 5 Properties of a Random Sample

Chapter 5 Properties of a Random Sample Lecture 6 o BST 63: Statstcal Theory I Ku Zhag, /0/008 Revew for the prevous lecture Cocepts: t-dstrbuto, F-dstrbuto Theorems: Dstrbutos of sample mea ad sample varace, relatoshp betwee sample mea ad sample

More information

Chapter 8. Inferences about More Than Two Population Central Values

Chapter 8. Inferences about More Than Two Population Central Values Chapter 8. Ifereces about More Tha Two Populato Cetral Values Case tudy: Effect of Tmg of the Treatmet of Port-We tas wth Lasers ) To vestgate whether treatmet at a youg age would yeld better results tha

More information

Lecture 2 - What are component and system reliability and how it can be improved?

Lecture 2 - What are component and system reliability and how it can be improved? Lecture 2 - What are compoet ad system relablty ad how t ca be mproved? Relablty s a measure of the qualty of the product over the log ru. The cocept of relablty s a exteded tme perod over whch the expected

More information

Lecture 7. Confidence Intervals and Hypothesis Tests in the Simple CLR Model

Lecture 7. Confidence Intervals and Hypothesis Tests in the Simple CLR Model Lecture 7. Cofdece Itervals ad Hypothess Tests the Smple CLR Model I lecture 6 we troduced the Classcal Lear Regresso (CLR) model that s the radom expermet of whch the data Y,,, K, are the outcomes. The

More information

Bayes Estimator for Exponential Distribution with Extension of Jeffery Prior Information

Bayes Estimator for Exponential Distribution with Extension of Jeffery Prior Information Malaysa Joural of Mathematcal Sceces (): 97- (9) Bayes Estmator for Expoetal Dstrbuto wth Exteso of Jeffery Pror Iformato Hadeel Salm Al-Kutub ad Noor Akma Ibrahm Isttute for Mathematcal Research, Uverst

More information

Lecture 3. Sampling, sampling distributions, and parameter estimation

Lecture 3. Sampling, sampling distributions, and parameter estimation Lecture 3 Samplg, samplg dstrbutos, ad parameter estmato Samplg Defto Populato s defed as the collecto of all the possble observatos of terest. The collecto of observatos we take from the populato s called

More information

2006 Jamie Trahan, Autar Kaw, Kevin Martin University of South Florida United States of America

2006 Jamie Trahan, Autar Kaw, Kevin Martin University of South Florida United States of America SOLUTION OF SYSTEMS OF SIMULTANEOUS LINEAR EQUATIONS Gauss-Sedel Method 006 Jame Traha, Autar Kaw, Kev Mart Uversty of South Florda Uted States of Amerca kaw@eg.usf.edu Itroducto Ths worksheet demostrates

More information

Part 4b Asymptotic Results for MRR2 using PRESS. Recall that the PRESS statistic is a special type of cross validation procedure (see Allen (1971))

Part 4b Asymptotic Results for MRR2 using PRESS. Recall that the PRESS statistic is a special type of cross validation procedure (see Allen (1971)) art 4b Asymptotc Results for MRR usg RESS Recall that the RESS statstc s a specal type of cross valdato procedure (see Alle (97)) partcular to the regresso problem ad volves fdg Y $,, the estmate at the

More information

Simulation Output Analysis

Simulation Output Analysis Smulato Output Aalyss Summary Examples Parameter Estmato Sample Mea ad Varace Pot ad Iterval Estmato ermatg ad o-ermatg Smulato Mea Square Errors Example: Sgle Server Queueg System x(t) S 4 S 4 S 3 S 5

More information

Introduction to local (nonparametric) density estimation. methods

Introduction to local (nonparametric) density estimation. methods Itroducto to local (oparametrc) desty estmato methods A slecture by Yu Lu for ECE 66 Sprg 014 1. Itroducto Ths slecture troduces two local desty estmato methods whch are Parze desty estmato ad k-earest

More information

ENGI 3423 Simple Linear Regression Page 12-01

ENGI 3423 Simple Linear Regression Page 12-01 ENGI 343 mple Lear Regresso Page - mple Lear Regresso ometmes a expermet s set up where the expermeter has cotrol over the values of oe or more varables X ad measures the resultg values of aother varable

More information

Comparing Different Estimators of three Parameters for Transmuted Weibull Distribution

Comparing Different Estimators of three Parameters for Transmuted Weibull Distribution Global Joural of Pure ad Appled Mathematcs. ISSN 0973-768 Volume 3, Number 9 (207), pp. 55-528 Research Ida Publcatos http://www.rpublcato.com Comparg Dfferet Estmators of three Parameters for Trasmuted

More information

Estimation of Stress- Strength Reliability model using finite mixture of exponential distributions

Estimation of Stress- Strength Reliability model using finite mixture of exponential distributions Iteratoal Joural of Computatoal Egeerg Research Vol, 0 Issue, Estmato of Stress- Stregth Relablty model usg fte mxture of expoetal dstrbutos K.Sadhya, T.S.Umamaheswar Departmet of Mathematcs, Lal Bhadur

More information

A New Family of Transformations for Lifetime Data

A New Family of Transformations for Lifetime Data Proceedgs of the World Cogress o Egeerg 4 Vol I, WCE 4, July - 4, 4, Lodo, U.K. A New Famly of Trasformatos for Lfetme Data Lakhaa Watthaacheewakul Abstract A famly of trasformatos s the oe of several

More information

CS286.2 Lecture 4: Dinur s Proof of the PCP Theorem

CS286.2 Lecture 4: Dinur s Proof of the PCP Theorem CS86. Lecture 4: Dur s Proof of the PCP Theorem Scrbe: Thom Bohdaowcz Prevously, we have prove a weak verso of the PCP theorem: NP PCP 1,1/ (r = poly, q = O(1)). Wth ths result we have the desred costat

More information

Feature Selection: Part 2. 1 Greedy Algorithms (continued from the last lecture)

Feature Selection: Part 2. 1 Greedy Algorithms (continued from the last lecture) CSE 546: Mache Learg Lecture 6 Feature Selecto: Part 2 Istructor: Sham Kakade Greedy Algorthms (cotued from the last lecture) There are varety of greedy algorthms ad umerous amg covetos for these algorthms.

More information

CHAPTER VI Statistical Analysis of Experimental Data

CHAPTER VI Statistical Analysis of Experimental Data Chapter VI Statstcal Aalyss of Expermetal Data CHAPTER VI Statstcal Aalyss of Expermetal Data Measuremets do ot lead to a uque value. Ths s a result of the multtude of errors (maly radom errors) that ca

More information

best estimate (mean) for X uncertainty or error in the measurement (systematic, random or statistical) best

best estimate (mean) for X uncertainty or error in the measurement (systematic, random or statistical) best Error Aalyss Preamble Wheever a measuremet s made, the result followg from that measuremet s always subject to ucertaty The ucertaty ca be reduced by makg several measuremets of the same quatty or by mprovg

More information

UNIVERSITY OF OSLO DEPARTMENT OF ECONOMICS

UNIVERSITY OF OSLO DEPARTMENT OF ECONOMICS UNIVERSITY OF OSLO DEPARTMENT OF ECONOMICS Exam: ECON430 Statstcs Date of exam: Frday, December 8, 07 Grades are gve: Jauary 4, 08 Tme for exam: 0900 am 00 oo The problem set covers 5 pages Resources allowed:

More information

Simple Linear Regression

Simple Linear Regression Statstcal Methods I (EST 75) Page 139 Smple Lear Regresso Smple regresso applcatos are used to ft a model descrbg a lear relatoshp betwee two varables. The aspects of least squares regresso ad correlato

More information

Chapter 4 Multiple Random Variables

Chapter 4 Multiple Random Variables Revew for the prevous lecture: Theorems ad Examples: How to obta the pmf (pdf) of U = g (, Y) ad V = g (, Y) Chapter 4 Multple Radom Varables Chapter 44 Herarchcal Models ad Mxture Dstrbutos Examples:

More information

Third handout: On the Gini Index

Third handout: On the Gini Index Thrd hadout: O the dex Corrado, a tala statstca, proposed (, 9, 96) to measure absolute equalt va the mea dfferece whch s defed as ( / ) where refers to the total umber of dvduals socet. Assume that. The

More information

Bayes (Naïve or not) Classifiers: Generative Approach

Bayes (Naïve or not) Classifiers: Generative Approach Logstc regresso Bayes (Naïve or ot) Classfers: Geeratve Approach What do we mea by Geeratve approach: Lear p(y), p(x y) ad the apply bayes rule to compute p(y x) for makg predctos Ths s essetally makg

More information

Bounds on the expected entropy and KL-divergence of sampled multinomial distributions. Brandon C. Roy

Bounds on the expected entropy and KL-divergence of sampled multinomial distributions. Brandon C. Roy Bouds o the expected etropy ad KL-dvergece of sampled multomal dstrbutos Brado C. Roy bcroy@meda.mt.edu Orgal: May 18, 2011 Revsed: Jue 6, 2011 Abstract Iformato theoretc quattes calculated from a sampled

More information

Unimodality Tests for Global Optimization of Single Variable Functions Using Statistical Methods

Unimodality Tests for Global Optimization of Single Variable Functions Using Statistical Methods Malaysa Umodalty Joural Tests of Mathematcal for Global Optmzato Sceces (): of 05 Sgle - 5 Varable (007) Fuctos Usg Statstcal Methods Umodalty Tests for Global Optmzato of Sgle Varable Fuctos Usg Statstcal

More information

{ }{ ( )} (, ) = ( ) ( ) ( ) Chapter 14 Exercises in Sampling Theory. Exercise 1 (Simple random sampling): Solution:

{ }{ ( )} (, ) = ( ) ( ) ( ) Chapter 14 Exercises in Sampling Theory. Exercise 1 (Simple random sampling): Solution: Chapter 4 Exercses Samplg Theory Exercse (Smple radom samplg: Let there be two correlated radom varables X ad A sample of sze s draw from a populato by smple radom samplg wthout replacemet The observed

More information

Lecture 02: Bounding tail distributions of a random variable

Lecture 02: Bounding tail distributions of a random variable CSCI-B609: A Theorst s Toolkt, Fall 206 Aug 25 Lecture 02: Boudg tal dstrbutos of a radom varable Lecturer: Yua Zhou Scrbe: Yua Xe & Yua Zhou Let us cosder the ubased co flps aga. I.e. let the outcome

More information

Analysis of Variance with Weibull Data

Analysis of Variance with Weibull Data Aalyss of Varace wth Webull Data Lahaa Watthaacheewaul Abstract I statstcal data aalyss by aalyss of varace, the usual basc assumptos are that the model s addtve ad the errors are radomly, depedetly, ad

More information

Lecture 9: Tolerant Testing

Lecture 9: Tolerant Testing Lecture 9: Tolerat Testg Dael Kae Scrbe: Sakeerth Rao Aprl 4, 07 Abstract I ths lecture we prove a quas lear lower boud o the umber of samples eeded to do tolerat testg for L dstace. Tolerat Testg We have

More information

Lecture 07: Poles and Zeros

Lecture 07: Poles and Zeros Lecture 07: Poles ad Zeros Defto of poles ad zeros The trasfer fucto provdes a bass for determg mportat system respose characterstcs wthout solvg the complete dfferetal equato. As defed, the trasfer fucto

More information

Discrete Mathematics and Probability Theory Fall 2016 Seshia and Walrand DIS 10b

Discrete Mathematics and Probability Theory Fall 2016 Seshia and Walrand DIS 10b CS 70 Dscrete Mathematcs ad Probablty Theory Fall 206 Sesha ad Walrad DIS 0b. Wll I Get My Package? Seaky delvery guy of some compay s out delverg packages to customers. Not oly does he had a radom package

More information

Chapter 3 Sampling For Proportions and Percentages

Chapter 3 Sampling For Proportions and Percentages Chapter 3 Samplg For Proportos ad Percetages I may stuatos, the characterstc uder study o whch the observatos are collected are qualtatve ature For example, the resposes of customers may marketg surveys

More information

ESS Line Fitting

ESS Line Fitting ESS 5 014 17. Le Fttg A very commo problem data aalyss s lookg for relatoshpetwee dfferet parameters ad fttg les or surfaces to data. The smplest example s fttg a straght le ad we wll dscuss that here

More information

Econometric Methods. Review of Estimation

Econometric Methods. Review of Estimation Ecoometrc Methods Revew of Estmato Estmatg the populato mea Radom samplg Pot ad terval estmators Lear estmators Ubased estmators Lear Ubased Estmators (LUEs) Effcecy (mmum varace) ad Best Lear Ubased Estmators

More information

The Mathematical Appendix

The Mathematical Appendix The Mathematcal Appedx Defto A: If ( Λ, Ω, where ( λ λ λ whch the probablty dstrbutos,,..., Defto A. uppose that ( Λ,,..., s a expermet type, the σ-algebra o λ λ λ are defed s deoted by ( (,,...,, σ Ω.

More information

Special Instructions / Useful Data

Special Instructions / Useful Data JAM 6 Set of all real umbers P A..d. B, p Posso Specal Istructos / Useful Data x,, :,,, x x Probablty of a evet A Idepedetly ad detcally dstrbuted Bomal dstrbuto wth parameters ad p Posso dstrbuto wth

More information

Lecture 3 Probability review (cont d)

Lecture 3 Probability review (cont d) STATS 00: Itroducto to Statstcal Iferece Autum 06 Lecture 3 Probablty revew (cot d) 3. Jot dstrbutos If radom varables X,..., X k are depedet, the ther dstrbuto may be specfed by specfyg the dvdual dstrbuto

More information

A New Method for Decision Making Based on Soft Matrix Theory

A New Method for Decision Making Based on Soft Matrix Theory Joural of Scetfc esearch & eports 3(5): 0-7, 04; rtcle o. JS.04.5.00 SCIENCEDOMIN teratoal www.scecedoma.org New Method for Decso Mag Based o Soft Matrx Theory Zhmg Zhag * College of Mathematcs ad Computer

More information

ENGI 4421 Propagation of Error Page 8-01

ENGI 4421 Propagation of Error Page 8-01 ENGI 441 Propagato of Error Page 8-01 Propagato of Error [Navd Chapter 3; ot Devore] Ay realstc measuremet procedure cotas error. Ay calculatos based o that measuremet wll therefore also cota a error.

More information

Cubic Nonpolynomial Spline Approach to the Solution of a Second Order Two-Point Boundary Value Problem

Cubic Nonpolynomial Spline Approach to the Solution of a Second Order Two-Point Boundary Value Problem Joural of Amerca Scece ;6( Cubc Nopolyomal Sple Approach to the Soluto of a Secod Order Two-Pot Boudary Value Problem W.K. Zahra, F.A. Abd El-Salam, A.A. El-Sabbagh ad Z.A. ZAk * Departmet of Egeerg athematcs

More information

ρ < 1 be five real numbers. The

ρ < 1 be five real numbers. The Lecture o BST 63: Statstcal Theory I Ku Zhag, /0/006 Revew for the prevous lecture Deftos: covarace, correlato Examples: How to calculate covarace ad correlato Theorems: propertes of correlato ad covarace

More information

Multi Objective Fuzzy Inventory Model with. Demand Dependent Unit Cost and Lead Time. Constraints A Karush Kuhn Tucker Conditions.

Multi Objective Fuzzy Inventory Model with. Demand Dependent Unit Cost and Lead Time. Constraints A Karush Kuhn Tucker Conditions. It. Joural of Math. Aalyss, Vol. 8, 204, o. 4, 87-93 HIKARI Ltd, www.m-hkar.com http://dx.do.org/0.2988/jma.204.30252 Mult Objectve Fuzzy Ivetory Model wth Demad Depedet Ut Cost ad Lead Tme Costrats A

More information

Derivation of 3-Point Block Method Formula for Solving First Order Stiff Ordinary Differential Equations

Derivation of 3-Point Block Method Formula for Solving First Order Stiff Ordinary Differential Equations Dervato of -Pot Block Method Formula for Solvg Frst Order Stff Ordary Dfferetal Equatos Kharul Hamd Kharul Auar, Kharl Iskadar Othma, Zara Bb Ibrahm Abstract Dervato of pot block method formula wth costat

More information

PTAS for Bin-Packing

PTAS for Bin-Packing CS 663: Patter Matchg Algorthms Scrbe: Che Jag /9/00. Itroducto PTAS for B-Packg The B-Packg problem s NP-hard. If we use approxmato algorthms, the B-Packg problem could be solved polyomal tme. For example,

More information

The Necessarily Efficient Point Method for Interval Molp Problems

The Necessarily Efficient Point Method for Interval Molp Problems ISS 6-69 Eglad K Joural of Iformato ad omputg Scece Vol. o. 9 pp. - The ecessarly Effcet Pot Method for Iterval Molp Problems Hassa Mshmast eh ad Marzeh Alezhad + Mathematcs Departmet versty of Ssta ad

More information

C-1: Aerodynamics of Airfoils 1 C-2: Aerodynamics of Airfoils 2 C-3: Panel Methods C-4: Thin Airfoil Theory

C-1: Aerodynamics of Airfoils 1 C-2: Aerodynamics of Airfoils 2 C-3: Panel Methods C-4: Thin Airfoil Theory ROAD MAP... AE301 Aerodyamcs I UNIT C: 2-D Arfols C-1: Aerodyamcs of Arfols 1 C-2: Aerodyamcs of Arfols 2 C-3: Pael Methods C-4: Th Arfol Theory AE301 Aerodyamcs I Ut C-3: Lst of Subects Problem Solutos?

More information

Chapter 8: Statistical Analysis of Simulated Data

Chapter 8: Statistical Analysis of Simulated Data Marquette Uversty MSCS600 Chapter 8: Statstcal Aalyss of Smulated Data Dael B. Rowe, Ph.D. Departmet of Mathematcs, Statstcs, ad Computer Scece Copyrght 08 by Marquette Uversty MSCS600 Ageda 8. The Sample

More information

X ε ) = 0, or equivalently, lim

X ε ) = 0, or equivalently, lim Revew for the prevous lecture Cocepts: order statstcs Theorems: Dstrbutos of order statstcs Examples: How to get the dstrbuto of order statstcs Chapter 5 Propertes of a Radom Sample Secto 55 Covergece

More information

CIS 800/002 The Algorithmic Foundations of Data Privacy October 13, Lecture 9. Database Update Algorithms: Multiplicative Weights

CIS 800/002 The Algorithmic Foundations of Data Privacy October 13, Lecture 9. Database Update Algorithms: Multiplicative Weights CIS 800/002 The Algorthmc Foudatos of Data Prvacy October 13, 2011 Lecturer: Aaro Roth Lecture 9 Scrbe: Aaro Roth Database Update Algorthms: Multplcatve Weghts We ll recall aga) some deftos from last tme:

More information

Non-uniform Turán-type problems

Non-uniform Turán-type problems Joural of Combatoral Theory, Seres A 111 2005 106 110 wwwelsevercomlocatecta No-uform Turá-type problems DhruvMubay 1, Y Zhao 2 Departmet of Mathematcs, Statstcs, ad Computer Scece, Uversty of Illos at

More information

BERNSTEIN COLLOCATION METHOD FOR SOLVING NONLINEAR DIFFERENTIAL EQUATIONS. Aysegul Akyuz Dascioglu and Nese Isler

BERNSTEIN COLLOCATION METHOD FOR SOLVING NONLINEAR DIFFERENTIAL EQUATIONS. Aysegul Akyuz Dascioglu and Nese Isler Mathematcal ad Computatoal Applcatos, Vol. 8, No. 3, pp. 293-300, 203 BERNSTEIN COLLOCATION METHOD FOR SOLVING NONLINEAR DIFFERENTIAL EQUATIONS Aysegul Ayuz Dascoglu ad Nese Isler Departmet of Mathematcs,

More information

Median as a Weighted Arithmetic Mean of All Sample Observations

Median as a Weighted Arithmetic Mean of All Sample Observations Meda as a Weghted Arthmetc Mea of All Sample Observatos SK Mshra Dept. of Ecoomcs NEHU, Shllog (Ida). Itroducto: Iumerably may textbooks Statstcs explctly meto that oe of the weakesses (or propertes) of

More information

( ) = ( ) ( ) Chapter 13 Asymptotic Theory and Stochastic Regressors. Stochastic regressors model

( ) = ( ) ( ) Chapter 13 Asymptotic Theory and Stochastic Regressors. Stochastic regressors model Chapter 3 Asmptotc Theor ad Stochastc Regressors The ature of eplaator varable s assumed to be o-stochastc or fed repeated samples a regresso aalss Such a assumpto s approprate for those epermets whch

More information

Research Article A New Iterative Method for Common Fixed Points of a Finite Family of Nonexpansive Mappings

Research Article A New Iterative Method for Common Fixed Points of a Finite Family of Nonexpansive Mappings Hdaw Publshg Corporato Iteratoal Joural of Mathematcs ad Mathematcal Sceces Volume 009, Artcle ID 391839, 9 pages do:10.1155/009/391839 Research Artcle A New Iteratve Method for Commo Fxed Pots of a Fte

More information

0/1 INTEGER PROGRAMMING AND SEMIDEFINTE PROGRAMMING

0/1 INTEGER PROGRAMMING AND SEMIDEFINTE PROGRAMMING CONVEX OPIMIZAION AND INERIOR POIN MEHODS FINAL PROJEC / INEGER PROGRAMMING AND SEMIDEFINE PROGRAMMING b Luca Buch ad Natala Vktorova CONENS:.Itroducto.Formulato.Applcato to Kapsack Problem 4.Cuttg Plaes

More information

Unsupervised Learning and Other Neural Networks

Unsupervised Learning and Other Neural Networks CSE 53 Soft Computg NOT PART OF THE FINAL Usupervsed Learg ad Other Neural Networs Itroducto Mture Destes ad Idetfablty ML Estmates Applcato to Normal Mtures Other Neural Networs Itroducto Prevously, all

More information

The internal structure of natural numbers, one method for the definition of large prime numbers, and a factorization test

The internal structure of natural numbers, one method for the definition of large prime numbers, and a factorization test Fal verso The teral structure of atural umbers oe method for the defto of large prme umbers ad a factorzato test Emmaul Maousos APM Isttute for the Advacemet of Physcs ad Mathematcs 3 Poulou str. 53 Athes

More information

22 Nonparametric Methods.

22 Nonparametric Methods. 22 oparametrc Methods. I parametrc models oe assumes apror that the dstrbutos have a specfc form wth oe or more ukow parameters ad oe tres to fd the best or atleast reasoably effcet procedures that aswer

More information

UNIT 2 SOLUTION OF ALGEBRAIC AND TRANSCENDENTAL EQUATIONS

UNIT 2 SOLUTION OF ALGEBRAIC AND TRANSCENDENTAL EQUATIONS Numercal Computg -I UNIT SOLUTION OF ALGEBRAIC AND TRANSCENDENTAL EQUATIONS Structure Page Nos..0 Itroducto 6. Objectves 7. Ital Approxmato to a Root 7. Bsecto Method 8.. Error Aalyss 9.4 Regula Fals Method

More information

8.1 Hashing Algorithms

8.1 Hashing Algorithms CS787: Advaced Algorthms Scrbe: Mayak Maheshwar, Chrs Hrchs Lecturer: Shuch Chawla Topc: Hashg ad NP-Completeess Date: September 21 2007 Prevously we looked at applcatos of radomzed algorthms, ad bega

More information

5 Short Proofs of Simplified Stirling s Approximation

5 Short Proofs of Simplified Stirling s Approximation 5 Short Proofs of Smplfed Strlg s Approxmato Ofr Gorodetsky, drtymaths.wordpress.com Jue, 20 0 Itroducto Strlg s approxmato s the followg (somewhat surprsg) approxmato of the factoral,, usg elemetary fuctos:

More information

ENGI 4421 Joint Probability Distributions Page Joint Probability Distributions [Navidi sections 2.5 and 2.6; Devore sections

ENGI 4421 Joint Probability Distributions Page Joint Probability Distributions [Navidi sections 2.5 and 2.6; Devore sections ENGI 441 Jot Probablty Dstrbutos Page 7-01 Jot Probablty Dstrbutos [Navd sectos.5 ad.6; Devore sectos 5.1-5.] The jot probablty mass fucto of two dscrete radom quattes, s, P ad p x y x y The margal probablty

More information

arxiv: v1 [math.st] 24 Oct 2016

arxiv: v1 [math.st] 24 Oct 2016 arxv:60.07554v [math.st] 24 Oct 206 Some Relatoshps ad Propertes of the Hypergeometrc Dstrbuto Peter H. Pesku, Departmet of Mathematcs ad Statstcs York Uversty, Toroto, Otaro M3J P3, Caada E-mal: pesku@pascal.math.yorku.ca

More information

Generating Multivariate Nonnormal Distribution Random Numbers Based on Copula Function

Generating Multivariate Nonnormal Distribution Random Numbers Based on Copula Function 7659, Eglad, UK Joural of Iformato ad Computg Scece Vol. 2, No. 3, 2007, pp. 9-96 Geeratg Multvarate Noormal Dstrbuto Radom Numbers Based o Copula Fucto Xaopg Hu +, Jam He ad Hogsheg Ly School of Ecoomcs

More information

MEASURES OF DISPERSION

MEASURES OF DISPERSION MEASURES OF DISPERSION Measure of Cetral Tedecy: Measures of Cetral Tedecy ad Dsperso ) Mathematcal Average: a) Arthmetc mea (A.M.) b) Geometrc mea (G.M.) c) Harmoc mea (H.M.) ) Averages of Posto: a) Meda

More information

Bootstrap Method for Testing of Equality of Several Coefficients of Variation

Bootstrap Method for Testing of Equality of Several Coefficients of Variation Cloud Publcatos Iteratoal Joural of Advaced Mathematcs ad Statstcs Volume, pp. -6, Artcle ID Sc- Research Artcle Ope Access Bootstrap Method for Testg of Equalty of Several Coeffcets of Varato Dr. Navee

More information

18.657: Mathematics of Machine Learning

18.657: Mathematics of Machine Learning 8.657: Mathematcs of Mache Learg Lecturer: Phlppe Rgollet Lecture 3 Scrbe: James Hrst Sep. 6, 205.5 Learg wth a fte dctoary Recall from the ed of last lecture our setup: We are workg wth a fte dctoary

More information

Application of Calibration Approach for Regression Coefficient Estimation under Two-stage Sampling Design

Application of Calibration Approach for Regression Coefficient Estimation under Two-stage Sampling Design Authors: Pradp Basak, Kaustav Adtya, Hukum Chadra ad U.C. Sud Applcato of Calbrato Approach for Regresso Coeffcet Estmato uder Two-stage Samplg Desg Pradp Basak, Kaustav Adtya, Hukum Chadra ad U.C. Sud

More information

TESTS BASED ON MAXIMUM LIKELIHOOD

TESTS BASED ON MAXIMUM LIKELIHOOD ESE 5 Toy E. Smth. The Basc Example. TESTS BASED ON MAXIMUM LIKELIHOOD To llustrate the propertes of maxmum lkelhood estmates ad tests, we cosder the smplest possble case of estmatg the mea of the ormal

More information

THE ROYAL STATISTICAL SOCIETY HIGHER CERTIFICATE

THE ROYAL STATISTICAL SOCIETY HIGHER CERTIFICATE THE ROYAL STATISTICAL SOCIETY 00 EXAMINATIONS SOLUTIONS HIGHER CERTIFICATE PAPER I STATISTICAL THEORY The Socety provdes these solutos to assst caddates preparg for the examatos future years ad for the

More information

This lecture and the next. Why Sorting? Sorting Algorithms so far. Why Sorting? (2) Selection Sort. Heap Sort. Heapsort

This lecture and the next. Why Sorting? Sorting Algorithms so far. Why Sorting? (2) Selection Sort. Heap Sort. Heapsort Ths lecture ad the ext Heapsort Heap data structure ad prorty queue ADT Qucksort a popular algorthm, very fast o average Why Sortg? Whe doubt, sort oe of the prcples of algorthm desg. Sortg used as a subroute

More information

COMPROMISE HYPERSPHERE FOR STOCHASTIC DOMINANCE MODEL

COMPROMISE HYPERSPHERE FOR STOCHASTIC DOMINANCE MODEL Sebasta Starz COMPROMISE HYPERSPHERE FOR STOCHASTIC DOMINANCE MODEL Abstract The am of the work s to preset a method of rakg a fte set of dscrete radom varables. The proposed method s based o two approaches:

More information

PGE 310: Formulation and Solution in Geosystems Engineering. Dr. Balhoff. Interpolation

PGE 310: Formulation and Solution in Geosystems Engineering. Dr. Balhoff. Interpolation PGE 30: Formulato ad Soluto Geosystems Egeerg Dr. Balhoff Iterpolato Numercal Methods wth MATLAB, Recktewald, Chapter 0 ad Numercal Methods for Egeers, Chapra ad Caale, 5 th Ed., Part Fve, Chapter 8 ad

More information

Chapter 4 (Part 1): Non-Parametric Classification (Sections ) Pattern Classification 4.3) Announcements

Chapter 4 (Part 1): Non-Parametric Classification (Sections ) Pattern Classification 4.3) Announcements Aoucemets No-Parametrc Desty Estmato Techques HW assged Most of ths lecture was o the blacboard. These sldes cover the same materal as preseted DHS Bometrcs CSE 90-a Lecture 7 CSE90a Fall 06 CSE90a Fall

More information

Algorithms Design & Analysis. Hash Tables

Algorithms Design & Analysis. Hash Tables Algorthms Desg & Aalyss Hash Tables Recap Lower boud Order statstcs 2 Today s topcs Drect-accessble table Hash tables Hash fuctos Uversal hashg Perfect Hashg Ope addressg 3 Symbol-table problem Symbol

More information

To use adaptive cluster sampling we must first make some definitions of the sampling universe:

To use adaptive cluster sampling we must first make some definitions of the sampling universe: 8.3 ADAPTIVE SAMPLING Most of the methods dscussed samplg theory are lmted to samplg desgs hch the selecto of the samples ca be doe before the survey, so that oe of the decsos about samplg deped ay ay

More information

Some Notes on the Probability Space of Statistical Surveys

Some Notes on the Probability Space of Statistical Surveys Metodološk zvezk, Vol. 7, No., 200, 7-2 ome Notes o the Probablty pace of tatstcal urveys George Petrakos Abstract Ths paper troduces a formal presetato of samplg process usg prcples ad cocepts from Probablty

More information

Rademacher Complexity. Examples

Rademacher Complexity. Examples Algorthmc Foudatos of Learg Lecture 3 Rademacher Complexty. Examples Lecturer: Patrck Rebesch Verso: October 16th 018 3.1 Itroducto I the last lecture we troduced the oto of Rademacher complexty ad showed

More information

Uniform asymptotical stability of almost periodic solution of a discrete multispecies Lotka-Volterra competition system

Uniform asymptotical stability of almost periodic solution of a discrete multispecies Lotka-Volterra competition system Iteratoal Joural of Egeerg ad Advaced Research Techology (IJEART) ISSN: 2454-9290, Volume-2, Issue-1, Jauary 2016 Uform asymptotcal stablty of almost perodc soluto of a dscrete multspeces Lotka-Volterra

More information

Bayes Interval Estimation for binomial proportion and difference of two binomial proportions with Simulation Study

Bayes Interval Estimation for binomial proportion and difference of two binomial proportions with Simulation Study IJIEST Iteratoal Joural of Iovatve Scece, Egeerg & Techology, Vol. Issue 5, July 04. Bayes Iterval Estmato for bomal proporto ad dfferece of two bomal proportos wth Smulato Study Masoud Gaj, Solmaz hlmad

More information

Likewise, properties of the optimal policy for equipment replacement & maintenance problems can be used to reduce the computation.

Likewise, properties of the optimal policy for equipment replacement & maintenance problems can be used to reduce the computation. Whe solvg a vetory repleshmet problem usg a MDP model, kowg that the optmal polcy s of the form (s,s) ca reduce the computatoal burde. That s, f t s optmal to replesh the vetory whe the vetory level s,

More information

Comparison of Dual to Ratio-Cum-Product Estimators of Population Mean

Comparison of Dual to Ratio-Cum-Product Estimators of Population Mean Research Joural of Mathematcal ad Statstcal Sceces ISS 30 6047 Vol. 1(), 5-1, ovember (013) Res. J. Mathematcal ad Statstcal Sc. Comparso of Dual to Rato-Cum-Product Estmators of Populato Mea Abstract

More information

Class 13,14 June 17, 19, 2015

Class 13,14 June 17, 19, 2015 Class 3,4 Jue 7, 9, 05 Pla for Class3,4:. Samplg dstrbuto of sample mea. The Cetral Lmt Theorem (CLT). Cofdece terval for ukow mea.. Samplg Dstrbuto for Sample mea. Methods used are based o CLT ( Cetral

More information

Strong Convergence of Weighted Averaged Approximants of Asymptotically Nonexpansive Mappings in Banach Spaces without Uniform Convexity

Strong Convergence of Weighted Averaged Approximants of Asymptotically Nonexpansive Mappings in Banach Spaces without Uniform Convexity BULLETIN of the MALAYSIAN MATHEMATICAL SCIENCES SOCIETY Bull. Malays. Math. Sc. Soc. () 7 (004), 5 35 Strog Covergece of Weghted Averaged Appromats of Asymptotcally Noepasve Mappgs Baach Spaces wthout

More information

The number of observed cases The number of parameters. ith case of the dichotomous dependent variable. the ith case of the jth parameter

The number of observed cases The number of parameters. ith case of the dichotomous dependent variable. the ith case of the jth parameter LOGISTIC REGRESSION Notato Model Logstc regresso regresses a dchotomous depedet varable o a set of depedet varables. Several methods are mplemeted for selectg the depedet varables. The followg otato s

More information

TRIANGULAR MEMBERSHIP FUNCTIONS FOR SOLVING SINGLE AND MULTIOBJECTIVE FUZZY LINEAR PROGRAMMING PROBLEM.

TRIANGULAR MEMBERSHIP FUNCTIONS FOR SOLVING SINGLE AND MULTIOBJECTIVE FUZZY LINEAR PROGRAMMING PROBLEM. Abbas Iraq Joural of SceceVol 53No 12012 Pp. 125-129 TRIANGULAR MEMBERSHIP FUNCTIONS FOR SOLVING SINGLE AND MULTIOBJECTIVE FUZZY LINEAR PROGRAMMING PROBLEM. Iraq Tarq Abbas Departemet of Mathematc College

More information

Multivariate Transformation of Variables and Maximum Likelihood Estimation

Multivariate Transformation of Variables and Maximum Likelihood Estimation Marquette Uversty Multvarate Trasformato of Varables ad Maxmum Lkelhood Estmato Dael B. Rowe, Ph.D. Assocate Professor Departmet of Mathematcs, Statstcs, ad Computer Scece Copyrght 03 by Marquette Uversty

More information

Statistics Descriptive and Inferential Statistics. Instructor: Daisuke Nagakura

Statistics Descriptive and Inferential Statistics. Instructor: Daisuke Nagakura Statstcs Descrptve ad Iferetal Statstcs Istructor: Dasuke Nagakura (agakura@z7.keo.jp) 1 Today s topc Today, I talk about two categores of statstcal aalyses, descrptve statstcs ad feretal statstcs, ad

More information

Aitken delta-squared generalized Juncgk-type iterative procedure

Aitken delta-squared generalized Juncgk-type iterative procedure Atke delta-squared geeralzed Jucgk-type teratve procedure M. De la Se Isttute of Research ad Developmet of Processes. Uversty of Basque Coutry Campus of Leoa (Bzkaa) PO Box. 644- Blbao, 488- Blbao. SPAIN

More information

ECONOMETRIC THEORY. MODULE VIII Lecture - 26 Heteroskedasticity

ECONOMETRIC THEORY. MODULE VIII Lecture - 26 Heteroskedasticity ECONOMETRIC THEORY MODULE VIII Lecture - 6 Heteroskedastcty Dr. Shalabh Departmet of Mathematcs ad Statstcs Ida Isttute of Techology Kapur . Breusch Paga test Ths test ca be appled whe the replcated data

More information

STA 105-M BASIC STATISTICS (This is a multiple choice paper.)

STA 105-M BASIC STATISTICS (This is a multiple choice paper.) DCDM BUSINESS SCHOOL September Mock Eamatos STA 0-M BASIC STATISTICS (Ths s a multple choce paper.) Tme: hours 0 mutes INSTRUCTIONS TO CANDIDATES Do ot ope ths questo paper utl you have bee told to do

More information

Chapter 14 Logistic Regression Models

Chapter 14 Logistic Regression Models Chapter 4 Logstc Regresso Models I the lear regresso model X β + ε, there are two types of varables explaatory varables X, X,, X k ad study varable y These varables ca be measured o a cotuous scale as

More information

CHAPTER 4 RADICAL EXPRESSIONS

CHAPTER 4 RADICAL EXPRESSIONS 6 CHAPTER RADICAL EXPRESSIONS. The th Root of a Real Number A real umber a s called the th root of a real umber b f Thus, for example: s a square root of sce. s also a square root of sce ( ). s a cube

More information

Analysis of Lagrange Interpolation Formula

Analysis of Lagrange Interpolation Formula P IJISET - Iteratoal Joural of Iovatve Scece, Egeerg & Techology, Vol. Issue, December 4. www.jset.com ISS 348 7968 Aalyss of Lagrage Iterpolato Formula Vjay Dahya PDepartmet of MathematcsMaharaja Surajmal

More information

A New Measure of Probabilistic Entropy. and its Properties

A New Measure of Probabilistic Entropy. and its Properties Appled Mathematcal Sceces, Vol. 4, 200, o. 28, 387-394 A New Measure of Probablstc Etropy ad ts Propertes Rajeesh Kumar Departmet of Mathematcs Kurukshetra Uversty Kurukshetra, Ida rajeesh_kuk@redffmal.com

More information