An Improved Genetic Algorithm with Average-Bound Crossover and Wavelet Mutation Operations

Size: px
Start display at page:

Download "An Improved Genetic Algorithm with Average-Bound Crossover and Wavelet Mutation Operations"

Transcription

1 An Improved Genetc Algorthm wth Average-Bound Crossover and Wavelet Mutaton Operatons S.H. Lng and F.H.F. Leung Centre for Multmeda Sgnal Processng, Department of Electronc and Informaton Engneerng, The Hong Kong Polytechnc Unversty, Hung Hom, Kowloon, Hong Kong Abstract: Ths paper presents a real-coded genetc algorthm (RCGA) wth new genetc operatons (crossover and mutaton). They are called the average-bound crossover (ABX) and wavelet mutaton (WM). By ntroducng the proposed genetc operatons, both the soluton qualty and stablty are better than the RCGA wth conventonal genetc operatons. A sute of benchmark test functons are used to evaluate the performance of the proposed algorthm. Applcaton examples on economc load dspatch and tunng an assocatve-memory neural network are used to show the performance of the proposed RCGA. Keywords: Crossover, mutaton, real-coded genetc algorthm, assocatve-memory neural network, and economc load dspatch. I. INTRODUCTION Genetc algorthm (GA) s one evolutonary computaton technque [9] that can tackle complex optmzaton problems [9, 7, 3]. It has been appled n dfferent areas such as fuzzy control [3-4], tunng of neural or neural fuzzy network [5-6], path plannng [], greenhouse clmate control [], economc load dspatch [, 7], etc. Tradtonal bnary GA [5, 9, 9, 5] has some drawbacks when applyng to multdmensonal and hgh-precson numercal problems. The stuaton can be mproved f GA n real numbers s used. Each chromosome s coded as a vector of floatng pont numbers that has the same length as the soluton vector. A large doman can thus be handled. Much research effort has been spent to mprove the performance of real-coded GA (RCGA). In general, RCGA nvolves three operatons: selecton, crossover and mutaton. The selecton operaton s used to select the chromosomes from the populaton wth respect to some probablty dstrbuton based on ftness values. The crossover operaton s used to combne the nformaton of the selected chromosomes (parents) and generate the offsprng. The mutaton

2 operaton s used to change the offsprng genes. Selecton schemes such as rank-based selecton, eltst strateges, steady-state electon and tournament selecton were reported [5]. Recently, dfferent crossover operatons for RCGA have been proposed to mprove the effcency of the algorthm. The extended ntermedate recombnaton (crossover) (EIX) was proposed by Mühlenben et. al. []. The genes (varables) of the offsprng are chosen somewhere between the genes of the parents. It s capable of producng any pont wthn a hypercube slghtly larger than that defned by the parents. The unmodal normal dstrbuton crossover (UNDX) was proposed by Ono et. al. [, ] for handlng multmodal functons and non-separablty problems. UNDX mxes the parental nformaton and shows a good searchng ablty. However, t changes the fundamental concept that the crossover operaton should combne the parents to generate offsprng, not mxng the parents. The blend crossover (BLX-α) was proposed by Eshelman et. al. [7], whch combnes the parents to reproduce offsprng. It shows a good searchng ablty for separable functons. However, BLX-α has dffculty n handlng non-separablty optmsaton problems. Also, the above crossover operatons are not sutable for optmsaton problems wth the optmal pont located near the doman boundary. For mutaton operatons, the unform mutaton and nonunform mutaton can be found [9, ]. The unform mutaton s to change the value of a randomly selected gene to a value between ts upper and lower bounds. The non-unform mutaton s capable of fne-tunng the parameters by ncreasng or decreasng the value of a randomly selected gene wth respect to a weghted random number. The weght s usually a monotonc decreasng functon of the number of teraton. In ths paper, new genetc operatons of crossover and mutaton are proposed. The crossover operaton s called the average-bound crossover (ABX), whch combnes the average crossover and bound crossover. The average crossover manpulates the genes of the selected parents, the mnmum, and the maxmum possble values of the genes. The bound crossover s capable of movng the offsprng near the doman boundary. On realzng the ABX operaton, the offsprng spreads over the doman so that a hgher chance of reachng the global optmum can be obtaned. The proposed mutaton operaton s called the wavelet mutaton (WM), whch apples the wavelet theory [3-4, 8] to realze the mutaton. Wavelet s a tool to model sesmc sgnals by combnng dlatons and translatons of a smple, oscllatory functon (mother wavelet) of a fnte duraton. The wavelet functon has two propertes: ) the functon ntegrates to zero, and ) t s square ntegrable, or equvalently has fnte energy. Thanks to the propertes of the wavelet, the convergence and soluton stablty are mproved. By ntroducng these genetc operatons, the RCGA performs more effcently and provdes a faster convergence than the RCGA wth conventonal genetc operatons n a sute of 8 benchmark test functons [6, 8, 4, 3]. In addton, the RCGA wth the proposed operatons gves smaller standard devatons of results,.e. the

3 soluton qualty of the RCGA wth the proposed operatons s more stable. An expermental study wll be made to evaluate the searchng ablty of the proposed mutaton. Also, the senstvty of the parameter n WM and the senstvty of the genes ntal range for the proposed RCGA to the searchng performance wll be dscussed. Applcaton examples on economc load dspatch and tunng an assocatve-memory neural network are also gven to show the performance of the proposed RCGA. Ths paper s organzed as follows. Secton II presents the operaton of the proposed genetc operatons. Expermental studes and analyss are dscussed n Secton III. 8 benchmark test functons wll be used to evaluate the performance of the proposed method. Applcaton examples on economc load dspatch and tunng an assocatve memory neural network are gven n Secton IV. A concluson wll be drawn n Secton V. II. AVERAGE-BOUND CROSSOVER AND WAVELET MUTATION FOR RCGA The Real-Coded Genetc Algorthm (RCGA) process [5, 9, 5] s shown n Fg.. Frst, a set of populaton of chromosomes P s created. Each chromosome p contans some genes (varables). Second, the chromosomes are evaluated by a defned ftness functon. The better chromosomes wll return hgher ftness functon values n ths process. Thrd, some of the chromosomes are selected to undergo genetc operatons for reproducton by the method of normalzed geometrc rankng []. Normalzed geometrc rankng s a selecton based on a nonstatonary penalty functon, whch s a functon of the generaton number. As the number of generaton ncreases, the penalty ncreases that puts more and more selectve pressure on the RCGA to fnd the feasble soluton. In general, a hgher-rank chromosome wll have a hgher chance to be selected. Fourth, genetc operatons of crossover are performed. The crossover operaton s manly for exchangng nformaton between two parents that are obtaned by the selecton operaton. In the crossover operaton, one of the parameters s the probablty of crossover p c whch gves the expected number p c pop _ sze (where pop _ sze s the number of chromosomes n the populaton) of chromosomes that undergo the crossover operaton n a generaton. We propose a new crossover operaton here. Frst, four chromosomes are generated (nstead of two chromosomes n the conventonal RCGA) from two selected parents. Second, the best two offsprng n terms of the ftness value wll be selected to replace ther parents. After the crossover operaton, the mutaton operaton follows. It operates wth a parameter called the probablty of mutaton ( p m ). The mutaton operaton s to change the genes of the chromosomes n the populaton such that the features nherted from ther parents can be changed. After gong through the mutaton operaton, 3

4 the new offsprng wll be evaluated usng the ftness functon. The new populaton wll be formed when the new offsprng replaces the chromosome wth the smallest ftness value. After the operatons of selecton, crossover and mutaton, a new populaton s generated. Ths new populaton wll repeat the same process. Such an teratve process wll be termnated when a defned condton s met. The detals about the proposed crossover and mutaton operatons are gven below. A. Average-bound crossover operaton The crossover operaton s manly for exchangng nformaton from the two parents, chromosomes p and p, obtaned n the selecton process. The two parents wll eventually produce two offsprng. The average-bound crossover (ABX) comprses two operatons: average crossover and bound crossover. ) Average crossover ( p p ) + o s = c ( p max + p mn )( w a ) + ( p + p ) w o a s = () c () ) Bound crossover ( p ) b o s = p ( w b ) + max w (3) 3 max, p c ( p ) b o s = p ( w b ) + mn w (4) 4 mn, p c where k = [ o k o k o k ] o, k =,, 3, 4. sc s s sno _ vars [ p p p p ] p =, =, ; =,,, no_vars, (5) no_vars paramn p para max max, (6) no _ vars [ para para para ] p =, (7) mn max max max no _ vars [ para para para ] p =, (8) mn mn no_vars denotes the number of varables to be tuned; maxmum values of mn para mn and p respectvely for all ;, [ ] a w b average crossover and bound crossover respectvely, max ( p ) para max are the mnmum and w denotes the user-defned weght for p denotes the vector wth each element obtaned by takng the maxmum between the correspondng element of p and p. For, 4

5 nstance, max( [ 3 ], [ 3 ] ) = [ 3 3]. Smlarly, mn( p ) the mnmum value. For nstance, mn( [ 3 ], [ 3 ] ) [ ] p, gves a vector by takng =. Among o to s c two wth the largest ftness values are used as the offsprng of the crossover operaton. These two offsprng are put back nto the populaton to replace ther parents. The ratonale behnd the ABX s that f the offsprng spreads over the doman, a hgher chance of reachng the global optmum can be obtaned. As seen from () to (4): The average crossover wll move the offsprng near the centre regon of the concerned doman (as w a n () o 4 s c, the approaches, approaches, o approaches ( p ) + p, whch s the average of the selected parents; and as w a s c o approaches p + p ), whch s the average of the doman boundary), c s ( max mn whle bound crossover wll move the offsprng near the doman boundary (as w b n (3) and (4) approaches, o s 3c and 4 s c o approaches p max and p mn respectvely). The result of the crossover depends on the values of the weghts w a and w b. Ther values depend on the optmsaton problem and are chosen by tral and error. Fg. shows an example ndcatng the relatonshp between the parents and the offsprng under dfferent values of the weghts. In ths fgure, the lne represents the doman of a gene. The end ponts of the lne represent the mnmum and maxmum values of the gene. The dot ( ) represents the parents and the crcle-dot ( ) represents the offsprng. The values n brackets represent the values of the genes under dfferent values of the weghts. For example, when p = and p = 4, referrng to () and (), the offsprng s c o and o should be equal s c to.5 and 3.75 respectvely when w =. 5. Accordng to () to (4), the offsprng s generated. We a can see how the offsprng spreads over the doman under dfferent values of w a and w b. Changng the value of the weght w wll change the characterstcs of the average crossover operaton. In a ths paper, the value of w a s arbtrarly set at.5. On the other hand, changng the value of the weght w wll change the characterstcs of the bound crossover operaton. b B. Wavelet mutaton operaton Before presentng the wavelet mutaton operaton, we frst dscuss the basc wavelet theory. ) Wavelet theory Certan sesmc sgnals can be modelled by combnng translatons and dlatons of an oscllatory functon wth a fnte duraton called a wavelet. A contnuous-tme functon ψ (x) s called a mother wavelet or wavelet f t satsfes the followng propertes: 5

6 Property : + ψ ( x) dx = (9) In other words, the total postve momentum of ψ (x) s equal to the total negatve momentum of ψ (x). Property : + ψ ( ) dx < () x where most of the energy n ψ (x) s confned to a fnte duraton and bounded. The Morlet wavelet (as shown n Fg.3) s an example mother wavelet, whch was proposed by Daubeches [4]: x / ( x) = e cos( 5x) ψ () The Morlet wavelet ntegrates to zero (Property ). Over 99% of the total energy of the functon s contaned n the nterval of.5 x. 5 (Property ). In order to control the magntude and the poston of ψ (x), we defne ψ ( ) as follows: a, b x x b ψ a, b ( x) = ψ () a a where a s the dlaton parameter and b s the translaton parameter. Notce that ( x) ψ, ( x) = ψ (3) As x ψ ( a, x) = ψ, (4) a a t follows that ψ ) s an ampltude-scaled verson of ψ (x). Fg. 4 shows dfferent dlatons of a, ( x the Morlet wavelet. The ampltude of ψ ) wll be scaled down as the dlaton parameter a a, ( x ncreases. Ths property s used to do the mutaton operaton n order to enhance the searchng performance. ) Wavelet mutaton The mutaton operaton s to change the genes of the chromosomes nherted from ther parents. In general, varous methods lke unform mutaton or non-unform mutaton [9, ] can be employed to realze the mutaton operaton. We propose a Wavelet Mutaton (WM) operaton based on the wavelet theory, whch exhbts a fne-tunng ablty. The detals of the operaton are as follows. Every gene of the chromosomes wll have a chance to mutate governed by a probablty of mutaton, p [ ] m, whch s defned by the user. Ths probablty gves an expected number ( p m pop _ sze no_vars) of genes that undergo the mutaton. For each gene, a random number 6

7 between and wll be generated such that f t s less than or equal to place on that gene whch s updated nstantly. If s = [ os, os, o ] chromosome and the element [ para, p m, the mutaton wll take o s the selected s no_ vars o s s randomly selected for mutaton (the value of s o s nsde mn, paramax ]), the resultng chromosome s gven by ˆ [ os oˆ s =,, s,, os no ] _ vars where,, no_vars, and ( paramax os ) ( o para ) o, os + δ f δ > oˆ s =, (5) os + δ s mn f δ δ = ψ a, ( ϕ) (6) δ = ϕ ψ a a By usng the Morlet wavelet n () as the mother wavelet, (7) δ = ϕ a e a ϕ cos 5 a where ϕ [.5,.5] s randomly generated. If δ s postve ( δ > ) approachng, the mutated (8) gene wll tend to the maxmum value of o. Conversely, when δ s negatve ( δ ) approachng s, the mutated gene wll tend to the mnmum value of searchng space for o s. A larger value of δ gves a larger o s. When δ s small, t gves a smaller searchng space for fne-tunng the gene. Referrng to Property of the wavelet, the sum of the postve δ s equal to the sum of the negatve δ when the number of samples s large and ϕ s randomly generated. That s, N N δ = for N, (9) where N s the number of samples. Hence, the overall postve mutaton and the overall negatve mutaton throughout the evoluton are nearly the same. Ths property gves better soluton stablty (smaller standard devaton of the soluton values upon many trals). As over 99% of the total energy of the mother wavelet functon s contaned n the nterval [.5,.5], ϕ can be generated from [.5,.5] randomly. The value of the dlaton parameter a can be set to vary wth the value of T τ n order to meet the fne-tunng purpose, where T s the total number of teraton and τ s the current number of teraton. In order to perform a local search when τ s large, the value of a should ncrease as T τ ncreases so as to 7

8 reduce the sgnfcance of the mutaton. Hence, a monotonc ncreasng functon governng a and τ s proposed as follows. T a = e ζ ln T + τ ( g ) ln( g ) where ζ s the shape parameter of the monotonc ncreasng functon, g s the upper lmt of the parameter a. In ths paper, g s set as. The effects of the varous values of the shape parameter ζ to a wth respect to T τ are shown n Fg. 5. The value of a s between and. τ Referrng to (8), the maxmum value of δ s when the random number of ϕ = and a= ( = ). T Then referrng to (5), the offsprng gene ( ) s s s () oˆ = o + para o = para. It ensures that a large search space for the mutated gene s gven. When the value T τ s near to, the value of a s τ so large that the maxmum value of δ wll become very small. For example, at =. 9 T max max and ζ =, the dlaton parameter a = 4. If the random value of ϕ s zero, the value of δ wll be equal to.58. Wth oˆ o +.58 ( para o ) gven for fne-tunng. s = s max s, a small searchng space for the mutated gene s C. Choosng the parameters We can regard the RCGA s seekng a balance between the exploraton of new regons and the explotaton of the already sampled regons n the search space. Ths balance, whch crtcally controls the performance of the RCGA, s governed by the rght choces of the control parameters: the probablty of crossover ( p ), the probablty of mutaton p ), the populaton sze (pop_sze), c the weghts of the proposed crossover ( w, w ) and the shape parameter ζ of WM. Some vews about these parameters are ncluded as follows: a b The probablty of crossover ( p c ) gves us an expected number ( p c pop_sze) of chromosomes whch undergo the crossover operaton n a generaton. When p c =, all ( m chromosomes n a generaton wll undergo the crossover operaton. Increasng the probablty of mutaton ( p m ) tends to transform the genetc search nto a random search. Ths probablty gves us an expected number ( p m pop_sze no_vars) of genes that undergo the mutaton. When p =, all genes wll mutate. The value of p m depends on the desred number of genes that undergo the mutaton operaton. m 8

9 Increasng the populaton sze wll ncrease the dversty of the search space, and reduce the probablty that GA wll prematurely converge to a local optmum. However, t also ncreases the tme requred for the populaton to converge to the optmal regon n the search space. Changng the value of the weght w a n the average-bounded crossover wll change the characterstcs of the average crossover operatons. It s chosen by tral and error, whch depends on the knd of the optmsaton problem. As the value of w a tends to, the offsprng tends to be the average of the selected parents. As the value of wa tends to, the offsprng tends to be the average of the doman boundary. For many optmsaton problems, the value of the weght w a can be set as.5. Changng the value of the weght w b n the average-bound crossover wll change the characterstcs of the bound crossover operatons. It s also chosen by tral and error, whch depends on the knd of the optmsaton problem. A value of w b approachng wll make the offsprng to be near the selected parents. As the value of w b tends to, the offsprng wll become near the doman boundary. Changng the parameter ζ wll change the characterstcs of the monotonc ncreasng functon of the wavelet mutaton. The dlaton parameter a wll take a value so as to perform fne-tunng faster as ζ s ncreasng. It s chosen by tral and error, whch depends on the knd of the optmsaton problem. When ζ becomes larger, the decreasng speed of the step sze (δ ) of the mutaton becomes faster. In general, f the optmsaton problem s smooth and symmetrc, the searchng algorthm s easer to fnd the soluton and process the fne-tunng n early teraton. Thus, a larger value of ζ can be used to ncrease the step sze of the early mutaton. More detals about the senstvty of ζ to WM wll be dscussed n the next secton. III. EXPERIMENTAL STUDIES AND ANALYSIS A. Benchmark test functon A sute of 8 benchmark test functons [6, 8, 4, 3] are used to test the performance of the RCGA wth the proposed genetc operatons. Many dfferent knds of optmzaton problems are covered by these benchmark test functons. They are dvded nto three categores: unmodal functons, multmodal functons wth only a few local mnma, and multmodal functons wth many local mnma. The 8 benchmark test functons are detaled n Appendx A. They can test 9

10 the searchng ablty of the proposed searchng algorthm comprehensvely. To avod the proposed crossover operaton ntroducng a strong bas to the optmal locaton at p + p ), the ranges ( max mn of the doman boundary for some test functons are set dfferent from those n [6, 8, 4, 3]. Functons f to f 7 are unmodal functons. Functons f 8 to f 3 are multmodal functons wth only a few local mnma. Functons f 4 to f 8 are multmodal functons wth many local mnma. B. Expermental setup The crossover operaton for comparson s the UNDXBXover, whch conssts of two publshed crossover operatons: Unmodal normal dstrbuton crossover (UNDX) [, ] and Blend crossover (BLX-α) [7]. The detals of these two crossovers are shown n Appendx B and Appendx C respectvely. The mutaton operaton for comparson s the non-unform mutaton (NUM) [9, ]. The detals of NUM are shown n Appendx D. The smulaton condtons are descrbed as follows. The shape parameter of NUM: It s chosen by tral and error through experments for good performance for all functons. The parameters ζ of WM: It s chosen by tral and error through experments for good performance for all functons. The weght of the ABX w a :.5 for all functons. The weght of the ABX w b :.5 for f to f 8 and f 5 to f 7 ;. for. f 9 to f 4, and f 8. Populaton sze:. Number of runs: 5. Selecton operaton: Normalzed geometrc rankng []. The probablty of selectng the best chromosome []:.8. Crossover operaton: For UNDX, the parameters β and µ are set at and.35 respectvely; for BLX-α, the parameter α s set at.336 [6]. Probablty of crossover p c :.8. Probablty of mutaton p m :.5 for f to f 6 and f 4 to f 8 ;.8 for f 7 to f 3. Intal populaton: It s generated unformly at random. In ths paper, RCGA wth Avergae-Bound Crossover and Wavelet Mutaton (), RCGA wth Avergae-Bound Crossover and Non-Unform Mutaton (), RCGA wth Unmodal Normal Dstrbuton and Blend Crossover and Wavelet Mutaton, (), and RCGA wth Unmodal Normal Dstrbuton and Blend crossover and Non-Unform Mutaton () are used to test the benchmark test functons.

11 C. Experment results. Unmodel Functons Functons f to f 6 are unmodal functons. The experment results n terms of the mean cost value, best cost value, standard devaton, and the t-test value for f to f 6 are tabulated n Table. I. The comparson between dfferent genetc operatons on f to f 6 s shown n Fg. 6. The t-test s a statstcal method to evaluate the sgnfcant dfference between two algorthms. The t- value wll be negatve f the frst algorthm s better than the second, and postve f t s poorer. When the t-value s smaller than.645 (degree of freedom = 49), there s a sgnfcant dfference between the two algorthms wth a 95% confdence level. Functon f s a sphere model whch s probably the most wdely used test functon. It s smooth and symmetrc. The performance on ths functon s a measure of the convergence rate of a searchng algorthm. For f, the results n terms of the mean and the best cost value of ABX wth WM or NUM are better than those of the correspondng UNDXBXover. Comparng ABX wth WM to UNDXBXover wth WM, the mean cost value s.5 tmes better. A much smaller standard devaton s gven by the, whch means the soluton s more stable. Comparng the mutaton operatons WM and NUM, the proposed WM s more effectve than NUM n term of the cost value and standard devaton. Both the soluton qualty and stablty offered by WM are better than those offered by NUM. In addton, the t value of.6 mples that the mproved genetc operatons (AveXover wth WM) are better than the conventonal genetc operatons (UNDXBXover wth NUM). In Fg. 6, ABX wth WM dsplays a faster convergence rate than UNDXBXover wth NUM thanks to ts better searchng ablty. It reaches approxmately. n around 5 tmes of teraton, whle t s about 3. for UNDXBXover wth NUM. Functon f s a generalzed Rosenbrock s functon whch s strongly non-separable and the optmum s located n a very narrow rdge. The tp of the rdge s very sharp, and t runs around a parabola. Algorthms that are unable to dscover good searchng drectons wll perform poorly n ths problem. The proposed genetc operatons (ABX wth WM) outperforms the UNDXBXover wth NUM. The t value s Although the best cost values on usng WM wth dfferent crossover operatons are a bt worse than those on usng NUM, the mean value, standard devaton and convergence rate offered by WM are better. Functon f 3 s a step functon that s a representatve of flat surfaces. Flat surfaces are obstacles for optmzaton algorthms because they do not gve any nformaton about the search drecton. Unless the algorthm has a varable step sze, t can get stuck n one of the flat surfaces. UNDXBXover performs poorly for f 3 because t manly searches n a small local neghbourhood, but the flat

12 surfaces do not gve any searchng drecton for UNDXBXover. On the other hand, the proposed ABX s good for f 3 because t can generate longer ump than UNDXBXover. Comparng WM to NUM wth UNDXBXover, the former also gves a better soluton. Functon f 4 s a quartc functon, whch s a smple unmodal functon padded wth nose. The Gaussan nose causes the algorthm never gettng the same value at the same pont. Many algorthms that do not do well n ths functon are due to the nosy data. In ths functon, the mean cost value, best cost value, standard devaton and the convergence rate brought by the proposed ABX and WM are sgnfcantly better than the conventonal genetc operatons. Functon f 5 s the Schwefel s problem., functon f 6 s the Schwefel s problem. and functon f 7 s the Eason s functon. In these problems, the performance on usng the proposed crossover ABX and mutaton WM s better than that on usng the UNDXBXover and NUM. The rapd convergence of the proposed genetc operatons shown n Fg. 6 supports our argument. In short, the proposed genetc operatons (ABX and WM) are good to tackle unmodal functons/problems when compared wth the conventonal genetc operatons (UNDXBXover and NUM). Both the soluton qualty and stablty are satsfactory.. Multmodel functons wth a few local mnma Functons f 8 to f 3 are multmodal functons wth only a few local mnma. The expermental results for f 8 to f 3 are tabulated n Table II. Fg. 7 shows the average values for f 8 to f 3. Among these functons, fve of them ( f 8, f - f 3 ) do not show statstcally sgnfcant dfferences for dfferent genetc operatons. They all reach or get near to the global optma. For the functon f 9, we obtan statstcally dfferent results from the proposed genetc operatons and the conventonal genetc operatons. The proposed ABX performs better than the UNDXBXover n terms of the mean, best value, standard devaton and the convergence rate. In addton, the results offered by WM are better than those by NUM n terms of the mean and the best cost values. Furthermore, WM gves a faster convergence rate. 3. Multmodel functons wth many local mnma Functons f 4 to f 8 are multmodal functons wth many local mnma, and the dmenson of each functon s comparatvely larger than that of f 8 to f 3. The dmenson of these functons s 3. The expermental results for f 4 to f 8 are tabulated n Table III. The comparson between dfferent genetc operatons s shown n Fg. 8. It can be seen from Table III that the mean results and the best results offered by the proposed genetc operatons (ABX and WM) are better than

13 those offered by the conventonal genetc operatons (UNDXBXover and NUM). Also, they have smaller standard devatons. Therefore, n terms of the soluton qualty and stablty, the proposed genetc operatons are better than the conventonal operatons. In addton, the t-test value of all functons s smaller than.645. Therefore, the proposed genetc operatons are sgnfcantly better than the conventonal operatons for solvng the optmzaton problems. From Fg. 8, we can see that the convergence rate offered by the proposed genetc operatons s better than that offered by the conventonal genetc operatons. D. The searchng ablty of wavelet mutaton In ths secton, we gve an analyss based on expermental results to llustrate that the searchng ablty of WM s better than that of NUM. The expermental settngs are the same as before, except the probablty of crossover s set at and the probablty of mutaton s set at. By usng ths settng, no chromosomes wll undergo the crossover operaton, and all genes n the populaton wll mutate under the mutaton operaton. Hence, the searchng ablty of the mutaton operaton can be evaluated. The expermental results on usng WM and NUM for f to f 8 (except f 3 and f 7 ) wthout crossover operaton are summarzed n Table IV. The comparson between WM and NUM s gven n Fg. 9 and Fg.. Functon f 3 and f 7 are not ncluded n ths experment because they do not perform well wth mutaton operaton only. As seen from Table IV, the average performance of WM s better than NUM. WM gves smaller standard devatons of results for all test functons than NUM, and hence the soluton stablty offered by WM s better. From Fg. 9 and Fg., the convergence of WM s found faster than that of NUM. In concluson, the searchng ablty of WM s better than NUM. E. Senstvty of the parameter for wavelet mutaton The mean cost values offered by WM usng dfferent shape parameter ζ for all test functons are tabulated n Table V. As can be seen from the table, all functons are tested by usng ζ =., ζ =.5, ζ =, ζ =, and ζ =5. If the optmzaton problem needs a more sgnfcant mutaton to reach the optmal pont, a smaller ζ should be gven. Conversely, f the RCGA needs to perform the fne-tunng faster, a larger ζ should be used. For example, f s a sphere model whch s smooth and symmetrc. Searchng algorthms are fast to ump to the area near the global optmum and then perform fne-tunng. Therefore, a larger ζ s set (ζ =5) so that the RCGA wll go to perform fne-tunng faster. On the other hand, ζ s set as. for f 3 when the mutaton operaton s playng a sgnfcant role at the later stage. In some cases, ζ s value s not very crtcal, e.g. f 3 and f. For f 3, the mean cost value for dfferent ζ s the same. We say that the best 3

14 performance s obtaned when ζ =.5 because the standard devaton of the RCGA for ζ =.5 s the smallest. However, n some cases, the parameter ζ s so senstve as to affect the performance of the searchng, e.g. f and f 6. In concluson, no formal method s avalable to choose the parameterζ ; t depends on the characterstcs of the optmzaton problems. F. Senstvty of the ntal range of varables Addtonal experments are carred out to test the senstvty of the ntal range of the varables to the RCGA wth the mproved genetc operatons. The settngs of these experments are exactly the same as before (secton III B). The experment results for f 4 are tabulated n Table VI. Fg. shows the results for dfferent genetc operatons on f 4. The ntal populaton s generated unformly at random n the ranges of.56 x 5. (twce the orgnal range), 6.4 x. 8 (5 tmes of the orgnal range),.8 x 5. 6 ( tmes of the orgnal range), and 5.6 x 5. ( tmes of the orgnal range), makng the average dstance to the global optmum ncreasngly large. The enlarged searchng space s expected to make the problem more dffcult to solve. As can be seen from the table and the fgures, the mean cost values offered by the proposed genetc operatons are better than those by the conventonal genetc operatons. From Fg., ABX and WM offer faster convergence than UNDXBXover and NUM. In addton, ABX and WM gve smaller standard devaton for all ntal ranges than UNDXBXover and NUM. Hence, the soluton qualty s more stable. Two more test functons are then used to test the senstvty to the ntal range of varables. The experment results for f 7 and f 6 are tabulated n Table VII to VIII respectvely. Fg. and Fg. 3 show the results for dfferent genetc operatons on f 7 and f 6 respectvely. In these tables and fgures, the results of the mproved genetc operatons n terms of the mean cost value, convergence rate, and standard devaton are better than those of the conventonal algorthms. IV. APPLICATION EXAMPLES Applcaton examples on economc load dspatch and tunng of assocatve memory are gven n ths secton. A. Economc load dspatch In a power system, mnmzng the operaton cost s mportant. Economc load dspatch (ELD) s a method to schedule power generator outputs wth respect to the load demands, and to operate a power system economcally. The nput-output characterstcs of modern generators are 4

15 nonlnear by nature because of the valve-pont loadngs and rate lmts. The problem of ELD s multmodal, dscontnuous and hghly nonlnear. RCGAs had been employed to solve the ELD problems [, 7].. Mathematc modellng of economc load dspatch wth valve-pont loadng The economc load dspatch wth valve-pont loadng problem can be formulated nto the followng obectve functon: Mn n = C ( P ) L, () where C ( P L ) s the operaton fuel cost of generator, and n denotes the number of generators. The problem s subect to a balance constrant and generatng capacty constrants as follows: D = P L n = P L P Loss, () L P P, =,,, n. (3) L, mn,max where D s the load demand, loss, P L,max and PL s the output power of the -th generator, P Loss s the transmsson P L,mn are the maxmum and mnmum output powers of the -th generator respectvely. The operaton fuel cost functon wth valve-pont loadngs of the generators s gven by, C ( P ) a P + b P + c + e sn( f ( P P ) L = L L L, mn L, (4) where a, b, and c are coeffcents of the cost curve of the -th generator, e and f are coeffcents of the valve-pont loadngs. (The generatng unts wth multvalve steam turbnes exhbt a great varaton n the fuel-cost functons. The valve-pont effects ntroduce rpples n the heat-rate curves.) RCGA can be used to solve the economc load dspatch problem. The chromosomes p s defned as follows: [ P P P P ] p, (5) = L L L3 L n From (), we have, n PL = D PL + P n Loss. (6) = In ths paper, the power loss s not consdered. Therefore, n P = D P. (7) Ln = L 5

16 To ensure f f PL falls wthn the range [ P ] n L, P n, mn Ln,max P = P + ( P P ), the followng condtons are consdered: L New L Ln,max Ln P L > P n L, (8) n,max PL = P n Ln,max ( P P ) PL = New PL Ln Ln,mn P L < P n L. (9) n,mn PL = P n Ln,mn It should be noted from (8) and (9) that f the value of PL s also outsde the constrant boundares. The exceedng porton of the power wll further be shared by other generators n order to make sure that all generators output power s wthn the safety range. Referrng to (), the ftness functon for ths ELD problem s defned as: n ftness = C ( PL ) =, (3) where C ( P L ) s defned n (4). The obectve s to maxmze the ftness functon (3).. Case Study The RCGA wth the proposed genetc operatons and the RCGA wth the conventonal genetc operatons are appled to a 4-generator system, whch was adopted as an example n []. The system s a very large one wth nonlneartes. The data of the unts for ths example wth valve-pont loadngs are tabulated n Table IX. The load demand (D) s 5MW. For comparson purpose, RCGA wth ABX and WM, RCGA wth ABX and NUM, RCGA wth UNDXBXover and WM, and RCGA wth UNDXBXover and NUM are used to solve the ELD problem. The populaton sze used for all RCGAs s. All the smulaton results are averaged ones out of 5 runs. For the proposed ABX, the parameters w a and w b are set at.5. For the UNDXBXover, the parameters β, µ, and α are set at,.35, and.336 respectvely. For the proposed mutaton WM, the parameter ζ s set at. For the NUM, the shape parameter s set at. The probabltes of crossover and mutaton for all approaches are set at.6 by tral and error. For all approaches, the number of teraton s. The statstcal results are shown n Table X and Fg. 4. It can be seen that the RCGA wth the proposed ABX and WM performs better than other RCGAs wth conventonal genetc operatons (UNDXBXover and NUM) n terms of cost, t value, and standard devaton. Both the soluton qualty and stablty are good. The average cost s $8.4 and the best (mnmum) cost s $ The optmal dspatch soluton s summarzed n Table XI. B. Tunng assocatve memory 6

17 Learnng or tranng s one of the mportant ssues of neural networks. The learnng process ams to fnd a set of optmal network parameters. The wdely-used gradent methods [8, 3], such as MRI, MRII, MRIII rules, and back-propagaton technques, adust the network parameters based on the gradent nformaton of the ftness functon n order to reduce the mean square error over all nput patterns. One maor weakness of the gradent methods s that the dervatve nformaton of the ftness functon has to be known, meanng that the ftness functon has to be contnuous. Also, the learnng process s easly trapped n a local optmum, especally when the problems are multmodal and the learnng rules are network structure dependent. To tackle ths problem, the real-code genetc algorthm (RCGA) [5, 9, 5], was proposed for the optmzaton problem n a large, complex, non-dfferentable and multmodal doman [9]. RCGA s a good tranng algorthm for neural or neural-fuzzy networks [5-6]. The same RCGA can be used to tran many dfferent networks regardless of whether they are feed-forward one, recurrent one, assocatve memory or of other structure types. Ths generally saves a lot of human efforts n developng tranng algorthms for dfferent types of networks. Assocatve memory s one type of neural network that maps ts nput vector nto tself. Thus, the desred output vector s ts nput vector. 5 nput vectors are used for the learnng. The functon of the assocatve memory s gven by: k ( t) = w z ( t) y, k =,,, (3) = k where z(t) s the nput vector and w k s the weght of the lnk between the -th nput and the k-th output. The obectve s to mnmze the mean square error (MSE), whch s defned as follows: MSE = 5 k = t= ( z ( t) y ( t) ) k 5 k The ntal range of the weght w k s from to. For comparson purpose, RCGA wth ABX and WM, RCGA wth ABX and NUM, RCGA wth UNDXBXover and WM, and RCGA wth UNDXBXover and NUM are used to solve ths problem. The populaton sze used for all RCGAs s. All the smulaton results are averaged ones out of 5 runs. For the proposed ABX, the parameters w a and w b are set at.5 and respectvely. For the UNDXBXover, the parameters β, µ, and α are set at,.35, and.336 respectvely. For the proposed mutaton WM, the parameter ζ s set at. For the NUM, the shape parameter s set at. The probabltes of crossover and mutaton for all approaches are set at.8 and. by tral and error. For all approaches, the number of teraton s. The expermental results are tabulated n Table XII, and the comparson between dfferent genetc operatons s shown n Fg. 5. As can be seen from the table, the mean and the best cost value offered by ABX and WM are better. In addton, the smaller standard (3) 7

18 devaton mples a more stable soluton. The t-value for ths functon s 4.67, whch s a relatvely large fgure. In short, the proposed RCGA s good for tunng assocatve memory. V. CONCLUSION An RCGA wth mproved genetc operatons (average-bound crossover and wavelet mutaton) has been presented. By usng the proposed crossover operaton, the offsprng spreads over the doman so that the probablty of reproducng good offsprng s ncreased. In the proposed mutaton operaton, the wavelet theory s appled. Thanks to the propertes of the wavelet, both the soluton qualty and stablty are mproved. A sut of benchmark test functons has been used to llustrate the merts of the mproved genetc operatons. Examples on economc load dspatch and tunng assocatve memory have also been gven. ACKNOWLEDGEMENT The work descrbed n ths paper was substantally supported by a grant from the Hong Kong Polytechnc Unversty (PhD Student Account Code: RG9T). REFERENCES [] Caponetto R., Fortuna L., Nunnar G., Occhpnt L., and Xbla M.G., Soft computng for greenhouse clmate control, IEEE Trans., Fuzzy Systems, vol. 8, no. 6, pp , Dec.. [] Chen P.H. and Chang H.C., Large-scale economc dspatch by genetc algorthm, IEEE Trans. Power Syst., vol., pp. 7-4, Feb [3] Daubeches I., The wavelet transform, tme-frequency localzaton and sgnal analyss, IEEE Trans. Informaton Theory, vol. 36, no.5, pp. 96-5, Sep. 99. [4] Daubeches I., Ten lectures on wavelets. Phladelpha, PA: Socety for Industrcal and Appled Mathematcs, 99. [5] Davs L., Handbook of genetc algorthms. NY: Van Nostrand Renhold, 99. [6] De Jong K.A., An analyss of the behavor of a class of genetc adaptve systems, Ph.D. Thess, Unversty of Mchgan, Ann Arbor, MI, 975. [7] Eshelman L.J. and Schaffer J.D., Real-coded genetc algorthms and nterval-schemata, Foundatons of Genetc Algorthms, pp. 87-, 993. [8] Goldsten A.A. and Prce I.F., On descent from local mnma, Math. Comput., vol. 5, no. 5, 97. [9] Holland J.H., Adaptaton n natural and artfcal systems. Unversty of Mchgan Press, Ann Arbor, MI,

19 [] Jones J. and Houck C., On the use of non-statonary penalty functons to solve constraned optmzaton problems wth genetc algorthm, n Proc. 994 Int. Symp. Evolutonary Computaton, Ordando, 994, pp [] Judette H. and Youlal H., Fuzzy dynamc path plannng usng genetc algorthms, Electroncs Letters, vol. 36, no. 4, pp , Feb.. [] Kta H., Ono I., and Kobayash S., Theoretcal analyss of the unmodal normal dstrbuton crossover for real-coded genetc algorthms, n Proc. of the Congress on Evolutonary Computatonal (CEC998), World Congress on Computatonal Intellgence (WCCI 998), May 4-9, 998, pp [3] Lam H.K., Leung F.H.F., and Tam P.K.S., Desgn and stablty analyss of fuzzy model based nonlnear controller for nonlnear systems usng genetc algorthm, IEEE Trans. Syst., Man and Cybern, Part B: Cybernetcs, vol. 33, no., pp. 5-57, Feb. 3. [4] Leung F.H.F., Lam H.K., Lng S.H., and Tam P.K.S., "Optmal and stable fuzzy controllers for nonlnear systems usng an mproved genetc algorthm," IEEE Trans. Industral Electroncs, vol. 5, no., pp.7-8, Feb. 4. [5] Leung F.H.F, Lam H.K., Lng S.H., and Tam P.K.S., "Tunng of the structure and parameters of neural network usng an mproved genetc algorthm," IEEE Trans. Neural Networks, vol.4, no., pp.79-88, Jan. 3. [6] Lng S.H., Leung F.H.F, Lam H.K., and Tam P.K.S., "Short-term electrc load forecastng based on a neural fuzzy network," IEEE Trans. Industral Electroncs, vol. 5, no. 6, pp.35-36, Dec. 3. [7] Lu B.D., Chen C.Y., and Tsao J.Y., Desgn of adaptve fuzzy logc controller based on lngustchedge concepts and genetc algorthms, IEEE Trans. Systems, Man and Cybernetcs, Part B, vol. 3 no., pp. 3-53, Feb.. [8] Mallat S.G., A theory for multresoluton sgnal decomposton: the wavelet representaton, IEEE Trans. Pattern analyss and machne ntellgence, vol., no.7, pp , Jul [9] Mchalewcz Z., Genetc Algorthm + Data Structures = Evoluton Programs, nd extended ed. Sprnger-Verlag, 994. [] Mühlenken H. and Schlerkamp-Voosen D., Predctve models for the breeder genetc algorthm I. contnuous parameter optmzaton, Evolutonary Computaton, vol., no., pp. 5-49, 993. [] Neubauer A., A theoretcal analyss of the non-unform mutaton operator for the modfed genetc algorthm, n Proc. IEEE Int. Conf. Evolutonary Computaton, 997, Indanapols, pp [] Ono I. and Kobayash S., A real-coded genetc algorthm for functon optmzaton usng unmodal normal dstrbuton crossover, n Proc. 7 th ICGA, 997, pp [3] Pham D.T. and Karaboga D., Intellgent Optmzaton Technques, Genetc Algorthms, Tabu Search, Smulated Annealng and Neural Networks. Sprnger,. [4] Schwefel H.P., Numercal optmzaton of computer models. Chchester, Wley & Sons, 98. [5] Srnvas M.and Patnak L.M., "Genetc algorthms: a survey," IEEE Computer, vol. 7, ssue 6, pp. 7-6, June

20 [6] Takahash M. and Kta H., A crossover operator usng ndependent component analyss for realcoded genetc algorthms, n Proc. of the Congress on Evolutonary Computaton, (CEC),, pp [7] Walter D.C. and Sheble G.B., Genetc algorthm soluton of economc dspatch wth valve pont loadng, IEEE Trans. Power Syst., vol. 8, pp.35-33, Aug [8] Wdrow B. and Lehr M.A., "3 years of adaptve neural networks: Perceptron, madalne, and backpropagaton," Proceedngs of the IEEE, vol. 78, no. 9, pp , Sept. 99. [9] Yao X., Evolvng artfcal networks, Proceedngs of the IEEE, vol. 87, no. 7, pp , 999. [3] Yao X. and Lu Y., Evolutonary programmng made faster, IEEE Trans. Evolutonary Computaton, vol. 3, no., pp. 8-, July 999. [3] Zurada J.M., Introducton to Artfcal Neural Systems. West Info Access, 99. Procedure of the RCGA begn τ // τ : teraton number Intalze P(τ) // P(τ) : populaton for teraton τ Evaluate f(p(τ)) // f(p(τ)) :ftness functon whle (not termnaton condton) do begn τ τ+ Select parents p and p from P(τ ) Perform crossover operaton wth p c Four chromosomes wll be generated Select the best two offsprng n terms of the ftness value Perform mutaton operaton wth p m Reproduce a new P(τ) Evaluate f(p(τ)) end end Fg.. RCGA process. ( ) (.5) (4) ( ) () p O O p s c s c paramn paramax w a = paramn ( ) (4) ( ) () p 3 O p O 4 sc s c paramax w b = paramn ( ) ( ) (.5) (3.3) (4) () p O O s p s c c paramax w a ( ) (.5) (3.75) (4) ( ) () w para mn a p O O para p max s s c c =.75 =.5 (.75) ( ) (4) (5.5) ( ) () para mn w b O p p para 3 max s c O s 4 c (.5) ( ) (4) (7) ( ) () para mn w para b O p p O max 3 4 s c sc =.75 =.5 ( ) ( ) (.5) (4) (4.38) () p O p O s c paramn s c paramax w a =.5 (.5) ( ) (4) (8.5) ( ) ) paramn O 3 sc p p O 4 sc ( paramax w b =.5 ( ) (.5) (4) (5) ( ) () w para mn para a p O p max s c O s c = ( ) ( ) (4) () ( ) () para mn w para b O p p max O s 3 s c 4 c = parent offsprng p O s c Fg.. Parents and offsprng under dfferent values of the weghts wa and w ( w, w =,.5,.5,.75 and.) b a b

21 ψ ( x) x Fg. 3. The Morlet Wavelet. a= a=5 a= a= a= a=5 a= a= Fg. 4. A Morlet wavelet dlated by dfferent values of the parameter a (x-axs: x, y-axs: ψ ).) a, ( x 4 a 3 ζ = 5 ζ = dlaton parameter ζ = ζ =.5 ζ = τ T Fg. 5. The effect of the shape parameter ζ to a wth respected to τ. T

22 ftness cost value ftness cost value value teraton number f teraton number f ftness cost value value 3 5 cost ftness value value UNDXXover+WM teraton number f teraton number f 4 cost value ftness value - UNDXXover+WM ftness cost value value UNDXXover+NUM teraton number f teraton number f cost value ftness value teraton number f 7 Fg. 6. Comparsons between dfferent genetc operatons for f to f 7. All results are averaged ones over 5 runs.

23 3 - ftness cost value cost value ftness value teraton number f teraton number f ftness cost value cost value ftness value teraton number -3.4 f teraton number -.6 f cost value ftness value cost ftness value teraton number f teraton number f 3 Fg. 7. Comparson between dfferent genetc operatons for f 8 to f 3. All results are averaged ones over 5 runs. 3

24 3 ftness cost value value 5 cost ftness value value teraton number teraton number f 4 f 5 cost value ftness value - - ftness cost value teraton number teraton number f 6 f ftness cost value value AveBXover+NUM AveBXover+WM teraton number f 8 Fg. 8. Comparson between dfferent genetc operatons for f 4 to f 8. All results are averaged ones over 5 runs. 4

25 cost ftness value value WM NUM ftness cost value 3 NUM - WM teraton number f teraton number f 3 ftness cost value NUM ftness cost value value WM NUM - WM teraton number f teraton number f ftness cost value WM NUM ftness cost value NUM teraton number f 6 WM teraton number f cost ftness value value ftness cost value value NUM NUM WM teraton number WM teraton number f 9 Fg. 9. Comparson between WM and NUM for f to f (except f 3 and f 7 ) wthout the crossover operaton. All results are averaged ones over 5 runs. f 5

26 cost ftness value value NUM cost ftness value value WM NUM teraton number -.6 f WM teraton number f cost ftness value ftness cost value value 4 NUM -.8 NUM WM teraton number f 3 3 WM teraton number f 4 3 NUM cost value ftness value WM ftness cost value - WM NUM teraton number f teraton number f cost value ftness value - WM NUM cost ftness value NUM teraton number f 7 - WM teraton number f 8 Fg.. Comparson between WM and NUM for f to f 8 wthout the crossover operaton. All results are averaged ones over 5 runs. 6

27 4 3 ftness cost value - cost value ftness value teraton number 5 (a) teraton number 6 (b) cost ftness value cost value ftness value 3 - ABXr+NUM teraton number (c) teraton number (d) Fg.. Comparson between dfferent genetc operatons for f 4 wth dfferent ntal ranges of varables. All results are averaged over 5 runs cost value ftness value ftness cost value teraton number (a) teraton number (b) cost value ftness value ftness cost value teraton number teraton number (c) (d) Fg.. Comparson between dfferent genetc operatons for f 7 wth dfferent ntal ranges of varables. All results are averaged ones over 5 runs. 7

28 4 6 3 cost ftness value value - ftness cost value value teraton number 6 (a) teraton number 6 (b) ftness cost value ftness cost value teraton number (c) teraton number (d) Fg. 3. Comparson between dfferent genetc operatons for f 6 wth dfferent ntal ranges of vrables. All results are averaged ones over 5 runs..3 x cost value ftness value teraton number Fg. 4. Comparsons between dfferent genetc operatons for ELD. All results are averaged ones over 5 runs. 8

The Study of Teaching-learning-based Optimization Algorithm

The Study of Teaching-learning-based Optimization Algorithm Advanced Scence and Technology Letters Vol. (AST 06), pp.05- http://dx.do.org/0.57/astl.06. The Study of Teachng-learnng-based Optmzaton Algorthm u Sun, Yan fu, Lele Kong, Haolang Q,, Helongang Insttute

More information

2003 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media,

2003 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, 003 IEEE. Personal use of ths materal s permtted. Permsson from IEEE must be obtaned for all other uses, n any current or future meda, ncludng reprntng/republshng ths materal for advertsng or promotonal

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

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

Design and Optimization of Fuzzy Controller for Inverse Pendulum System Using Genetic Algorithm

Design and Optimization of Fuzzy Controller for Inverse Pendulum System Using Genetic Algorithm Desgn and Optmzaton of Fuzzy Controller for Inverse Pendulum System Usng Genetc Algorthm H. Mehraban A. Ashoor Unversty of Tehran Unversty of Tehran h.mehraban@ece.ut.ac.r a.ashoor@ece.ut.ac.r Abstract:

More information

Solving of Single-objective Problems based on a Modified Multiple-crossover Genetic Algorithm: Test Function Study

Solving of Single-objective Problems based on a Modified Multiple-crossover Genetic Algorithm: Test Function Study Internatonal Conference on Systems, Sgnal Processng and Electroncs Engneerng (ICSSEE'0 December 6-7, 0 Duba (UAE Solvng of Sngle-objectve Problems based on a Modfed Multple-crossover Genetc Algorthm: Test

More information

VQ widely used in coding speech, image, and video

VQ widely used in coding speech, image, and video at Scalar quantzers are specal cases of vector quantzers (VQ): they are constraned to look at one sample at a tme (memoryless) VQ does not have such constrant better RD perfomance expected Source codng

More information

EEE 241: Linear Systems

EEE 241: Linear Systems EEE : Lnear Systems Summary #: Backpropagaton BACKPROPAGATION The perceptron rule as well as the Wdrow Hoff learnng were desgned to tran sngle layer networks. They suffer from the same dsadvantage: they

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

Using Immune Genetic Algorithm to Optimize BP Neural Network and Its Application Peng-fei LIU1,Qun-tai SHEN1 and Jun ZHI2,*

Using Immune Genetic Algorithm to Optimize BP Neural Network and Its Application Peng-fei LIU1,Qun-tai SHEN1 and Jun ZHI2,* Advances n Computer Scence Research (ACRS), volume 54 Internatonal Conference on Computer Networks and Communcaton Technology (CNCT206) Usng Immune Genetc Algorthm to Optmze BP Neural Network and Its Applcaton

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

Solving Nonlinear Differential Equations by a Neural Network Method

Solving Nonlinear Differential Equations by a Neural Network Method Solvng Nonlnear Dfferental Equatons by a Neural Network Method Luce P. Aarts and Peter Van der Veer Delft Unversty of Technology, Faculty of Cvlengneerng and Geoscences, Secton of Cvlengneerng Informatcs,

More information

Chapter 5. Solution of System of Linear Equations. Module No. 6. Solution of Inconsistent and Ill Conditioned Systems

Chapter 5. Solution of System of Linear Equations. Module No. 6. Solution of Inconsistent and Ill Conditioned Systems Numercal Analyss by Dr. Anta Pal Assstant Professor Department of Mathematcs Natonal Insttute of Technology Durgapur Durgapur-713209 emal: anta.bue@gmal.com 1 . Chapter 5 Soluton of System of Lnear Equatons

More information

Chapter 2 Real-Coded Adaptive Range Genetic Algorithm

Chapter 2 Real-Coded Adaptive Range Genetic Algorithm Chapter Real-Coded Adaptve Range Genetc Algorthm.. Introducton Fndng a global optmum n the contnuous doman s challengng for Genetc Algorthms (GAs. Tradtonal GAs use the bnary representaton that evenly

More information

EEL 6266 Power System Operation and Control. Chapter 3 Economic Dispatch Using Dynamic Programming

EEL 6266 Power System Operation and Control. Chapter 3 Economic Dispatch Using Dynamic Programming EEL 6266 Power System Operaton and Control Chapter 3 Economc Dspatch Usng Dynamc Programmng Pecewse Lnear Cost Functons Common practce many utltes prefer to represent ther generator cost functons as sngle-

More information

Some modelling aspects for the Matlab implementation of MMA

Some modelling aspects for the Matlab implementation of MMA Some modellng aspects for the Matlab mplementaton of MMA Krster Svanberg krlle@math.kth.se Optmzaton and Systems Theory Department of Mathematcs KTH, SE 10044 Stockholm September 2004 1. Consdered optmzaton

More information

A New Evolutionary Computation Based Approach for Learning Bayesian Network

A New Evolutionary Computation Based Approach for Learning Bayesian Network Avalable onlne at www.scencedrect.com Proceda Engneerng 15 (2011) 4026 4030 Advanced n Control Engneerng and Informaton Scence A New Evolutonary Computaton Based Approach for Learnng Bayesan Network Yungang

More information

MMA and GCMMA two methods for nonlinear optimization

MMA and GCMMA two methods for nonlinear optimization MMA and GCMMA two methods for nonlnear optmzaton Krster Svanberg Optmzaton and Systems Theory, KTH, Stockholm, Sweden. krlle@math.kth.se Ths note descrbes the algorthms used n the author s 2007 mplementatons

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

The Convergence Speed of Single- And Multi-Objective Immune Algorithm Based Optimization Problems

The Convergence Speed of Single- And Multi-Objective Immune Algorithm Based Optimization Problems The Convergence Speed of Sngle- And Mult-Obectve Immune Algorthm Based Optmzaton Problems Mohammed Abo-Zahhad Faculty of Engneerng, Electrcal and Electroncs Engneerng Department, Assut Unversty, Assut,

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

Generalized Linear Methods

Generalized Linear Methods Generalzed Lnear Methods 1 Introducton In the Ensemble Methods the general dea s that usng a combnaton of several weak learner one could make a better learner. More formally, assume that we have a set

More information

Particle Swarm Optimization with Adaptive Mutation in Local Best of Particles

Particle Swarm Optimization with Adaptive Mutation in Local Best of Particles 1 Internatonal Congress on Informatcs, Envronment, Energy and Applcatons-IEEA 1 IPCSIT vol.38 (1) (1) IACSIT Press, Sngapore Partcle Swarm Optmzaton wth Adaptve Mutaton n Local Best of Partcles Nanda ulal

More information

2016 Wiley. Study Session 2: Ethical and Professional Standards Application

2016 Wiley. Study Session 2: Ethical and Professional Standards Application 6 Wley Study Sesson : Ethcal and Professonal Standards Applcaton LESSON : CORRECTION ANALYSIS Readng 9: Correlaton and Regresson LOS 9a: Calculate and nterpret a sample covarance and a sample correlaton

More information

One-sided finite-difference approximations suitable for use with Richardson extrapolation

One-sided finite-difference approximations suitable for use with Richardson extrapolation Journal of Computatonal Physcs 219 (2006) 13 20 Short note One-sded fnte-dfference approxmatons sutable for use wth Rchardson extrapolaton Kumar Rahul, S.N. Bhattacharyya * Department of Mechancal Engneerng,

More information

The Minimum Universal Cost Flow in an Infeasible Flow Network

The Minimum Universal Cost Flow in an Infeasible Flow Network Journal of Scences, Islamc Republc of Iran 17(2): 175-180 (2006) Unversty of Tehran, ISSN 1016-1104 http://jscencesutacr The Mnmum Unversal Cost Flow n an Infeasble Flow Network H Saleh Fathabad * M Bagheran

More information

NON-CENTRAL 7-POINT FORMULA IN THE METHOD OF LINES FOR PARABOLIC AND BURGERS' EQUATIONS

NON-CENTRAL 7-POINT FORMULA IN THE METHOD OF LINES FOR PARABOLIC AND BURGERS' EQUATIONS IJRRAS 8 (3 September 011 www.arpapress.com/volumes/vol8issue3/ijrras_8_3_08.pdf NON-CENTRAL 7-POINT FORMULA IN THE METHOD OF LINES FOR PARABOLIC AND BURGERS' EQUATIONS H.O. Bakodah Dept. of Mathematc

More information

arxiv: v1 [math.oc] 3 Aug 2010

arxiv: v1 [math.oc] 3 Aug 2010 arxv:1008.0549v1 math.oc] 3 Aug 2010 Test Problems n Optmzaton Xn-She Yang Department of Engneerng, Unversty of Cambrdge, Cambrdge CB2 1PZ, UK Abstract Test functons are mportant to valdate new optmzaton

More information

LINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity

LINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity LINEAR REGRESSION ANALYSIS MODULE IX Lecture - 31 Multcollnearty Dr. Shalabh Department of Mathematcs and Statstcs Indan Insttute of Technology Kanpur 6. Rdge regresson The OLSE s the best lnear unbased

More information

Evolutionary Computational Techniques to Solve Economic Load Dispatch Problem Considering Generator Operating Constraints

Evolutionary Computational Techniques to Solve Economic Load Dispatch Problem Considering Generator Operating Constraints Internatonal Journal of Engneerng Research and Applcatons (IJERA) ISSN: 48-96 Natonal Conference On Advances n Energy and Power Control Engneerng (AEPCE-K1) Evolutonary Computatonal Technques to Solve

More information

CHAPTER-5 INFORMATION MEASURE OF FUZZY MATRIX AND FUZZY BINARY RELATION

CHAPTER-5 INFORMATION MEASURE OF FUZZY MATRIX AND FUZZY BINARY RELATION CAPTER- INFORMATION MEASURE OF FUZZY MATRI AN FUZZY BINARY RELATION Introducton The basc concept of the fuzz matr theor s ver smple and can be appled to socal and natural stuatons A branch of fuzz matr

More information

Single-Facility Scheduling over Long Time Horizons by Logic-based Benders Decomposition

Single-Facility Scheduling over Long Time Horizons by Logic-based Benders Decomposition Sngle-Faclty Schedulng over Long Tme Horzons by Logc-based Benders Decomposton Elvn Coban and J. N. Hooker Tepper School of Busness, Carnege Mellon Unversty ecoban@andrew.cmu.edu, john@hooker.tepper.cmu.edu

More information

Week3, Chapter 4. Position and Displacement. Motion in Two Dimensions. Instantaneous Velocity. Average Velocity

Week3, Chapter 4. Position and Displacement. Motion in Two Dimensions. Instantaneous Velocity. Average Velocity Week3, Chapter 4 Moton n Two Dmensons Lecture Quz A partcle confned to moton along the x axs moves wth constant acceleraton from x =.0 m to x = 8.0 m durng a 1-s tme nterval. The velocty of the partcle

More information

Winter 2008 CS567 Stochastic Linear/Integer Programming Guest Lecturer: Xu, Huan

Winter 2008 CS567 Stochastic Linear/Integer Programming Guest Lecturer: Xu, Huan Wnter 2008 CS567 Stochastc Lnear/Integer Programmng Guest Lecturer: Xu, Huan Class 2: More Modelng Examples 1 Capacty Expanson Capacty expanson models optmal choces of the tmng and levels of nvestments

More information

A PROBABILITY-DRIVEN SEARCH ALGORITHM FOR SOLVING MULTI-OBJECTIVE OPTIMIZATION PROBLEMS

A PROBABILITY-DRIVEN SEARCH ALGORITHM FOR SOLVING MULTI-OBJECTIVE OPTIMIZATION PROBLEMS HCMC Unversty of Pedagogy Thong Nguyen Huu et al. A PROBABILITY-DRIVEN SEARCH ALGORITHM FOR SOLVING MULTI-OBJECTIVE OPTIMIZATION PROBLEMS Thong Nguyen Huu and Hao Tran Van Department of mathematcs-nformaton,

More information

Numerical Heat and Mass Transfer

Numerical Heat and Mass Transfer Master degree n Mechancal Engneerng Numercal Heat and Mass Transfer 06-Fnte-Dfference Method (One-dmensonal, steady state heat conducton) Fausto Arpno f.arpno@uncas.t Introducton Why we use models and

More information

829. An adaptive method for inertia force identification in cantilever under moving mass

829. An adaptive method for inertia force identification in cantilever under moving mass 89. An adaptve method for nerta force dentfcaton n cantlever under movng mass Qang Chen 1, Mnzhuo Wang, Hao Yan 3, Haonan Ye 4, Guola Yang 5 1,, 3, 4 Department of Control and System Engneerng, Nanng Unversty,

More information

Supporting Information

Supporting Information Supportng Informaton The neural network f n Eq. 1 s gven by: f x l = ReLU W atom x l + b atom, 2 where ReLU s the element-wse rectfed lnear unt, 21.e., ReLUx = max0, x, W atom R d d s the weght matrx to

More information

Appendix B: Resampling Algorithms

Appendix B: Resampling Algorithms 407 Appendx B: Resamplng Algorthms A common problem of all partcle flters s the degeneracy of weghts, whch conssts of the unbounded ncrease of the varance of the mportance weghts ω [ ] of the partcles

More information

Chapter Newton s Method

Chapter Newton s Method Chapter 9. Newton s Method After readng ths chapter, you should be able to:. Understand how Newton s method s dfferent from the Golden Secton Search method. Understand how Newton s method works 3. Solve

More information

A new Approach for Solving Linear Ordinary Differential Equations

A new Approach for Solving Linear Ordinary Differential Equations , ISSN 974-57X (Onlne), ISSN 974-5718 (Prnt), Vol. ; Issue No. 1; Year 14, Copyrght 13-14 by CESER PUBLICATIONS A new Approach for Solvng Lnear Ordnary Dfferental Equatons Fawz Abdelwahd Department of

More information

CSci 6974 and ECSE 6966 Math. Tech. for Vision, Graphics and Robotics Lecture 21, April 17, 2006 Estimating A Plane Homography

CSci 6974 and ECSE 6966 Math. Tech. for Vision, Graphics and Robotics Lecture 21, April 17, 2006 Estimating A Plane Homography CSc 6974 and ECSE 6966 Math. Tech. for Vson, Graphcs and Robotcs Lecture 21, Aprl 17, 2006 Estmatng A Plane Homography Overvew We contnue wth a dscusson of the major ssues, usng estmaton of plane projectve

More information

Chapter - 2. Distribution System Power Flow Analysis

Chapter - 2. Distribution System Power Flow Analysis Chapter - 2 Dstrbuton System Power Flow Analyss CHAPTER - 2 Radal Dstrbuton System Load Flow 2.1 Introducton Load flow s an mportant tool [66] for analyzng electrcal power system network performance. Load

More information

A Bayes Algorithm for the Multitask Pattern Recognition Problem Direct Approach

A Bayes Algorithm for the Multitask Pattern Recognition Problem Direct Approach A Bayes Algorthm for the Multtask Pattern Recognton Problem Drect Approach Edward Puchala Wroclaw Unversty of Technology, Char of Systems and Computer etworks, Wybrzeze Wyspanskego 7, 50-370 Wroclaw, Poland

More information

Chapter 3 Describing Data Using Numerical Measures

Chapter 3 Describing Data Using Numerical Measures Chapter 3 Student Lecture Notes 3-1 Chapter 3 Descrbng Data Usng Numercal Measures Fall 2006 Fundamentals of Busness Statstcs 1 Chapter Goals To establsh the usefulness of summary measures of data. The

More information

Annexes. EC.1. Cycle-base move illustration. EC.2. Problem Instances

Annexes. EC.1. Cycle-base move illustration. EC.2. Problem Instances ec Annexes Ths Annex frst llustrates a cycle-based move n the dynamc-block generaton tabu search. It then dsplays the characterstcs of the nstance sets, followed by detaled results of the parametercalbraton

More information

LOW BIAS INTEGRATED PATH ESTIMATORS. James M. Calvin

LOW BIAS INTEGRATED PATH ESTIMATORS. James M. Calvin Proceedngs of the 007 Wnter Smulaton Conference S G Henderson, B Bller, M-H Hseh, J Shortle, J D Tew, and R R Barton, eds LOW BIAS INTEGRATED PATH ESTIMATORS James M Calvn Department of Computer Scence

More information

Self-adaptive Differential Evolution Algorithm for Constrained Real-Parameter Optimization

Self-adaptive Differential Evolution Algorithm for Constrained Real-Parameter Optimization 26 IEEE Congress on Evolutonary Computaton Sheraton Vancouver Wall Centre Hotel, Vancouver, BC, Canada July 16-21, 26 Self-adaptve Dfferental Evoluton Algorthm for Constraned Real-Parameter Optmzaton V.

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

Comparative Studies of Law of Conservation of Energy. and Law Clusters of Conservation of Generalized Energy

Comparative Studies of Law of Conservation of Energy. and Law Clusters of Conservation of Generalized Energy Comparatve Studes of Law of Conservaton of Energy and Law Clusters of Conservaton of Generalzed Energy No.3 of Comparatve Physcs Seres Papers Fu Yuhua (CNOOC Research Insttute, E-mal:fuyh1945@sna.com)

More information

A Fast Computer Aided Design Method for Filters

A Fast Computer Aided Design Method for Filters 2017 Asa-Pacfc Engneerng and Technology Conference (APETC 2017) ISBN: 978-1-60595-443-1 A Fast Computer Aded Desgn Method for Flters Gang L ABSTRACT *Ths paper presents a fast computer aded desgn method

More information

Constrained Evolutionary Programming Approaches to Power System Economic Dispatch

Constrained Evolutionary Programming Approaches to Power System Economic Dispatch Proceedngs of the 6th WSEAS Int. Conf. on EVOLUTIONARY COMPUTING, Lsbon, Portugal, June 16-18, 2005 (pp160-166) Constraned Evolutonary Programmng Approaches to Power System Economc Dspatch K. Shant Swarup

More information

Simulated Power of the Discrete Cramér-von Mises Goodness-of-Fit Tests

Simulated Power of the Discrete Cramér-von Mises Goodness-of-Fit Tests Smulated of the Cramér-von Mses Goodness-of-Ft Tests Steele, M., Chaselng, J. and 3 Hurst, C. School of Mathematcal and Physcal Scences, James Cook Unversty, Australan School of Envronmental Studes, Grffth

More information

Thin-Walled Structures Group

Thin-Walled Structures Group Thn-Walled Structures Group JOHNS HOPKINS UNIVERSITY RESEARCH REPORT Towards optmzaton of CFS beam-column ndustry sectons TWG-RR02-12 Y. Shfferaw July 2012 1 Ths report was prepared ndependently, but was

More information

Comparison of the Population Variance Estimators. of 2-Parameter Exponential Distribution Based on. Multiple Criteria Decision Making Method

Comparison of the Population Variance Estimators. of 2-Parameter Exponential Distribution Based on. Multiple Criteria Decision Making Method Appled Mathematcal Scences, Vol. 7, 0, no. 47, 07-0 HIARI Ltd, www.m-hkar.com Comparson of the Populaton Varance Estmators of -Parameter Exponental Dstrbuton Based on Multple Crtera Decson Makng Method

More information

Markov Chain Monte Carlo (MCMC), Gibbs Sampling, Metropolis Algorithms, and Simulated Annealing Bioinformatics Course Supplement

Markov Chain Monte Carlo (MCMC), Gibbs Sampling, Metropolis Algorithms, and Simulated Annealing Bioinformatics Course Supplement Markov Chan Monte Carlo MCMC, Gbbs Samplng, Metropols Algorthms, and Smulated Annealng 2001 Bonformatcs Course Supplement SNU Bontellgence Lab http://bsnuackr/ Outlne! Markov Chan Monte Carlo MCMC! Metropols-Hastngs

More information

Interactive Bi-Level Multi-Objective Integer. Non-linear Programming Problem

Interactive Bi-Level Multi-Objective Integer. Non-linear Programming Problem Appled Mathematcal Scences Vol 5 0 no 65 3 33 Interactve B-Level Mult-Objectve Integer Non-lnear Programmng Problem O E Emam Department of Informaton Systems aculty of Computer Scence and nformaton Helwan

More information

Wavelet chaotic neural networks and their application to continuous function optimization

Wavelet chaotic neural networks and their application to continuous function optimization Vol., No.3, 04-09 (009) do:0.436/ns.009.307 Natural Scence Wavelet chaotc neural networks and ther applcaton to contnuous functon optmzaton Ja-Ha Zhang, Yao-Qun Xu College of Electrcal and Automatc Engneerng,

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

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

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

COMPARISON OF SOME RELIABILITY CHARACTERISTICS BETWEEN REDUNDANT SYSTEMS REQUIRING SUPPORTING UNITS FOR THEIR OPERATIONS

COMPARISON OF SOME RELIABILITY CHARACTERISTICS BETWEEN REDUNDANT SYSTEMS REQUIRING SUPPORTING UNITS FOR THEIR OPERATIONS Avalable onlne at http://sck.org J. Math. Comput. Sc. 3 (3), No., 6-3 ISSN: 97-537 COMPARISON OF SOME RELIABILITY CHARACTERISTICS BETWEEN REDUNDANT SYSTEMS REQUIRING SUPPORTING UNITS FOR THEIR OPERATIONS

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

10-701/ Machine Learning, Fall 2005 Homework 3

10-701/ Machine Learning, Fall 2005 Homework 3 10-701/15-781 Machne Learnng, Fall 2005 Homework 3 Out: 10/20/05 Due: begnnng of the class 11/01/05 Instructons Contact questons-10701@autonlaborg for queston Problem 1 Regresson and Cross-valdaton [40

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

BOOTSTRAP METHOD FOR TESTING OF EQUALITY OF SEVERAL MEANS. M. Krishna Reddy, B. Naveen Kumar and Y. Ramu

BOOTSTRAP METHOD FOR TESTING OF EQUALITY OF SEVERAL MEANS. M. Krishna Reddy, B. Naveen Kumar and Y. Ramu BOOTSTRAP METHOD FOR TESTING OF EQUALITY OF SEVERAL MEANS M. Krshna Reddy, B. Naveen Kumar and Y. Ramu Department of Statstcs, Osmana Unversty, Hyderabad -500 007, Inda. nanbyrozu@gmal.com, ramu0@gmal.com

More information

A DNA Coding Scheme for Searching Stable Solutions

A DNA Coding Scheme for Searching Stable Solutions A DNA odng Scheme for Searchng Stable Solutons Intaek Km, HeSong Lan, and Hwan Il Kang 2 Department of ommuncaton Eng., Myongj Unversty, 449-728, Yongn, South Korea kt@mju.ac.kr, hslan@hotmal.net 2 Department

More information

HYBRID FUZZY MULTI-OBJECTIVE EVOLUTIONARY ALGORITHM: A NOVEL PARETO-OPTIMIZATION TECHNIQUE

HYBRID FUZZY MULTI-OBJECTIVE EVOLUTIONARY ALGORITHM: A NOVEL PARETO-OPTIMIZATION TECHNIQUE Internatonal Journal of Fuzzy Logc Systems (IJFLS) Vol.2, No., February 22 HYBRID FUZZY MULTI-OBJECTIVE EVOLUTIONARY ALGORITHM: A NOVEL PARETO-OPTIMIZATION TECHNIQUE Amt Saraswat and Ashsh San 2 Department

More information

x = , so that calculated

x = , so that calculated Stat 4, secton Sngle Factor ANOVA notes by Tm Plachowsk n chapter 8 we conducted hypothess tests n whch we compared a sngle sample s mean or proporton to some hypotheszed value Chapter 9 expanded ths to

More information

Research Article Green s Theorem for Sign Data

Research Article Green s Theorem for Sign Data Internatonal Scholarly Research Network ISRN Appled Mathematcs Volume 2012, Artcle ID 539359, 10 pages do:10.5402/2012/539359 Research Artcle Green s Theorem for Sgn Data Lous M. Houston The Unversty of

More information

Determining Transmission Losses Penalty Factor Using Adaptive Neuro Fuzzy Inference System (ANFIS) For Economic Dispatch Application

Determining Transmission Losses Penalty Factor Using Adaptive Neuro Fuzzy Inference System (ANFIS) For Economic Dispatch Application 7 Determnng Transmsson Losses Penalty Factor Usng Adaptve Neuro Fuzzy Inference System (ANFIS) For Economc Dspatch Applcaton Rony Seto Wbowo Maurdh Hery Purnomo Dod Prastanto Electrcal Engneerng Department,

More information

Hongyi Miao, College of Science, Nanjing Forestry University, Nanjing ,China. (Received 20 June 2013, accepted 11 March 2014) I)ϕ (k)

Hongyi Miao, College of Science, Nanjing Forestry University, Nanjing ,China. (Received 20 June 2013, accepted 11 March 2014) I)ϕ (k) ISSN 1749-3889 (prnt), 1749-3897 (onlne) Internatonal Journal of Nonlnear Scence Vol.17(2014) No.2,pp.188-192 Modfed Block Jacob-Davdson Method for Solvng Large Sparse Egenproblems Hongy Mao, College of

More information

Assortment Optimization under MNL

Assortment Optimization under MNL Assortment Optmzaton under MNL Haotan Song Aprl 30, 2017 1 Introducton The assortment optmzaton problem ams to fnd the revenue-maxmzng assortment of products to offer when the prces of products are fxed.

More information

Pulse Coded Modulation

Pulse Coded Modulation Pulse Coded Modulaton PCM (Pulse Coded Modulaton) s a voce codng technque defned by the ITU-T G.711 standard and t s used n dgtal telephony to encode the voce sgnal. The frst step n the analog to dgtal

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

Semi-supervised Classification with Active Query Selection

Semi-supervised Classification with Active Query Selection Sem-supervsed Classfcaton wth Actve Query Selecton Jao Wang and Swe Luo School of Computer and Informaton Technology, Beng Jaotong Unversty, Beng 00044, Chna Wangjao088@63.com Abstract. Labeled samples

More information

Simultaneous Optimization of Berth Allocation, Quay Crane Assignment and Quay Crane Scheduling Problems in Container Terminals

Simultaneous Optimization of Berth Allocation, Quay Crane Assignment and Quay Crane Scheduling Problems in Container Terminals Smultaneous Optmzaton of Berth Allocaton, Quay Crane Assgnment and Quay Crane Schedulng Problems n Contaner Termnals Necat Aras, Yavuz Türkoğulları, Z. Caner Taşkın, Kuban Altınel Abstract In ths work,

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

Feature Selection: Part 1

Feature Selection: Part 1 CSE 546: Machne Learnng Lecture 5 Feature Selecton: Part 1 Instructor: Sham Kakade 1 Regresson n the hgh dmensonal settng How do we learn when the number of features d s greater than the sample sze n?

More information

An Extended Hybrid Genetic Algorithm for Exploring a Large Search Space

An Extended Hybrid Genetic Algorithm for Exploring a Large Search Space 2nd Internatonal Conference on Autonomous Robots and Agents Abstract An Extended Hybrd Genetc Algorthm for Explorng a Large Search Space Hong Zhang and Masum Ishkawa Graduate School of L.S.S.E., Kyushu

More information

Statistics for Economics & Business

Statistics for Economics & Business Statstcs for Economcs & Busness Smple Lnear Regresson Learnng Objectves In ths chapter, you learn: How to use regresson analyss to predct the value of a dependent varable based on an ndependent varable

More information

De-noising Method Based on Kernel Adaptive Filtering for Telemetry Vibration Signal of the Vehicle Test Kejun ZENG

De-noising Method Based on Kernel Adaptive Filtering for Telemetry Vibration Signal of the Vehicle Test Kejun ZENG 6th Internatonal Conference on Mechatroncs, Materals, Botechnology and Envronment (ICMMBE 6) De-nosng Method Based on Kernel Adaptve Flterng for elemetry Vbraton Sgnal of the Vehcle est Kejun ZEG PLA 955

More information

Operating conditions of a mine fan under conditions of variable resistance

Operating conditions of a mine fan under conditions of variable resistance Paper No. 11 ISMS 216 Operatng condtons of a mne fan under condtons of varable resstance Zhang Ynghua a, Chen L a, b, Huang Zhan a, *, Gao Yukun a a State Key Laboratory of Hgh-Effcent Mnng and Safety

More information

Resource Allocation with a Budget Constraint for Computing Independent Tasks in the Cloud

Resource Allocation with a Budget Constraint for Computing Independent Tasks in the Cloud Resource Allocaton wth a Budget Constrant for Computng Independent Tasks n the Cloud Wemng Sh and Bo Hong School of Electrcal and Computer Engneerng Georga Insttute of Technology, USA 2nd IEEE Internatonal

More information

Neural networks. Nuno Vasconcelos ECE Department, UCSD

Neural networks. Nuno Vasconcelos ECE Department, UCSD Neural networs Nuno Vasconcelos ECE Department, UCSD Classfcaton a classfcaton problem has two types of varables e.g. X - vector of observatons (features) n the world Y - state (class) of the world x X

More information

COEFFICIENT DIAGRAM: A NOVEL TOOL IN POLYNOMIAL CONTROLLER DESIGN

COEFFICIENT DIAGRAM: A NOVEL TOOL IN POLYNOMIAL CONTROLLER DESIGN Int. J. Chem. Sc.: (4), 04, 645654 ISSN 097768X www.sadgurupublcatons.com COEFFICIENT DIAGRAM: A NOVEL TOOL IN POLYNOMIAL CONTROLLER DESIGN R. GOVINDARASU a, R. PARTHIBAN a and P. K. BHABA b* a Department

More information

Logistic Regression. CAP 5610: Machine Learning Instructor: Guo-Jun QI

Logistic Regression. CAP 5610: Machine Learning Instructor: Guo-Jun QI Logstc Regresson CAP 561: achne Learnng Instructor: Guo-Jun QI Bayes Classfer: A Generatve model odel the posteror dstrbuton P(Y X) Estmate class-condtonal dstrbuton P(X Y) for each Y Estmate pror dstrbuton

More information

DUE: WEDS FEB 21ST 2018

DUE: WEDS FEB 21ST 2018 HOMEWORK # 1: FINITE DIFFERENCES IN ONE DIMENSION DUE: WEDS FEB 21ST 2018 1. Theory Beam bendng s a classcal engneerng analyss. The tradtonal soluton technque makes smplfyng assumptons such as a constant

More information

Multilayer Perceptrons and Backpropagation. Perceptrons. Recap: Perceptrons. Informatics 1 CG: Lecture 6. Mirella Lapata

Multilayer Perceptrons and Backpropagation. Perceptrons. Recap: Perceptrons. Informatics 1 CG: Lecture 6. Mirella Lapata Multlayer Perceptrons and Informatcs CG: Lecture 6 Mrella Lapata School of Informatcs Unversty of Ednburgh mlap@nf.ed.ac.uk Readng: Kevn Gurney s Introducton to Neural Networks, Chapters 5 6.5 January,

More information

FFT Based Spectrum Analysis of Three Phase Signals in Park (d-q) Plane

FFT Based Spectrum Analysis of Three Phase Signals in Park (d-q) Plane Proceedngs of the 00 Internatonal Conference on Industral Engneerng and Operatons Management Dhaka, Bangladesh, January 9 0, 00 FFT Based Spectrum Analyss of Three Phase Sgnals n Park (d-q) Plane Anuradha

More information

Copyright 2017 by Taylor Enterprises, Inc., All Rights Reserved. Adjusted Control Limits for P Charts. Dr. Wayne A. Taylor

Copyright 2017 by Taylor Enterprises, Inc., All Rights Reserved. Adjusted Control Limits for P Charts. Dr. Wayne A. Taylor Taylor Enterprses, Inc. Control Lmts for P Charts Copyrght 2017 by Taylor Enterprses, Inc., All Rghts Reserved. Control Lmts for P Charts Dr. Wayne A. Taylor Abstract: P charts are used for count data

More information

CHAPTER 7 STOCHASTIC ECONOMIC EMISSION DISPATCH-MODELED USING WEIGHTING METHOD

CHAPTER 7 STOCHASTIC ECONOMIC EMISSION DISPATCH-MODELED USING WEIGHTING METHOD 90 CHAPTER 7 STOCHASTIC ECOOMIC EMISSIO DISPATCH-MODELED USIG WEIGHTIG METHOD 7.1 ITRODUCTIO early 70% of electrc power produced n the world s by means of thermal plants. Thermal power statons are the

More information

Research on Route guidance of logistic scheduling problem under fuzzy time window

Research on Route guidance of logistic scheduling problem under fuzzy time window Advanced Scence and Technology Letters, pp.21-30 http://dx.do.org/10.14257/astl.2014.78.05 Research on Route gudance of logstc schedulng problem under fuzzy tme wndow Yuqang Chen 1, Janlan Guo 2 * Department

More information

Lecture 12: Discrete Laplacian

Lecture 12: Discrete Laplacian Lecture 12: Dscrete Laplacan Scrbe: Tanye Lu Our goal s to come up wth a dscrete verson of Laplacan operator for trangulated surfaces, so that we can use t n practce to solve related problems We are mostly

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

Lecture 10 Support Vector Machines. Oct

Lecture 10 Support Vector Machines. Oct Lecture 10 Support Vector Machnes Oct - 20-2008 Lnear Separators Whch of the lnear separators s optmal? Concept of Margn Recall that n Perceptron, we learned that the convergence rate of the Perceptron

More information

Energy Conversion and Management

Energy Conversion and Management Energy Converson and Management 49 (2008) 3036 3042 Contents lsts avalable at ScenceDrect Energy Converson and Management ournal homepage: www.elsever.com/locate/enconman Modfed dfferental evoluton algorthm

More information

An Adaptive Learning Particle Swarm Optimizer for Function Optimization

An Adaptive Learning Particle Swarm Optimizer for Function Optimization An Adaptve Learnng Partcle Swarm Optmzer for Functon Optmzaton Changhe L and Shengxang Yang Abstract Tradtonal partcle swarm optmzaton (PSO) suffers from the premature convergence problem, whch usually

More information

An Improved multiple fractal algorithm

An Improved multiple fractal algorithm Advanced Scence and Technology Letters Vol.31 (MulGraB 213), pp.184-188 http://dx.do.org/1.1427/astl.213.31.41 An Improved multple fractal algorthm Yun Ln, Xaochu Xu, Jnfeng Pang College of Informaton

More information

Fuzzy Boundaries of Sample Selection Model

Fuzzy Boundaries of Sample Selection Model Proceedngs of the 9th WSES Internatonal Conference on ppled Mathematcs, Istanbul, Turkey, May 7-9, 006 (pp309-34) Fuzzy Boundares of Sample Selecton Model L. MUHMD SFIIH, NTON BDULBSH KMIL, M. T. BU OSMN

More information