Appl .Mat h.J.Chi ne s eUni v.Se r .B 2005,20(4):491498 A DERI VATI VEFREEALGORI THM FOR UNCONSTRAI NED OPTI MI ZATI ON 1,2 Pe ngYe hui 1 Li uZhe nha i .I Abs t r ac t nt hi spa pe rahybr i da l gor i t hm whi c hc ombi ne st hepa t t e r ns e a r c hme t hoda ndt he ge ne t i ca l gor i t hm f orunc ons t r a i ne dopt i mi z a t i oni spr e s e nt e d.Thea l gor i t hm i sade t e r mi ni s t i c pa t t e r ns e a r c ha l gor i t hm,buti nt hes e a r c hs t e pofpa t t e r ns e a r c ha l gor i t hm,t het r i a lpoi nt sa r e pr oduc e dbyawa yl i ket hege ne t i ca l gor i t hm.Ate a c hi t e r a t e,byr e dupl i c a t i on,c r os s ove ra nd .The mut a t i on,af i ni t es e tofpoi nt sc a nbeus e d.I nt he or y,t hea l gor i t hm i sgl oba l l yc onve r ge nt mos ts t i ri st henume r i c a lr e s ul t ss howi ng t ha ti tc a nf i nd t hegl oba lmi ni mi z e rf ors ome ,whi pr obl e ms c hot he rpa t t e r ns e a r c ha l gor i t hmsdon' tbe a r. &’ I (t r )*+c t , )( Cons i de rt hef ol l owi ngopt i mi z at i onpr obl e m: mi n-(.) ./ Rn, n (1) 1 whe r e-:R 0R i sc ont i nuous l ydi f f e r e nt i abl e. 3 1he nt hee 2pl i c i ti nf or mat i on about -(.) i sunavai l abl eorunt r us t wor t hy,t he de r i vat i vef r e eopt i mi z at i on me t hodsar et hebe s tc hoi c e st os ol vepr obl e m(1).Pat t e r n ,(s s e ar c h me t hodi sane f f i c i e ntde r i vat i vef r e eme t hodf oropt i mi z at i onpr obl e ms e e41, 25).Ate ac hi t e r at e,pat t e r ns e ar c hal gor i t hmspe r f or mt hes e ar c hs t e pandt hepol ls t e pt o 35).A pat f i nd abe t t e rt r i alpoi ntt hant hec ur r e ntone,(s e e41t e r ns e ar c h al gor i t hm ne e dn' ti nc l udet hes e ar c hs t e p,butt hee f f i c i e nts e ar c hs t e pc ani mpr ovet hepe r f or manc e. Fore 2ampl e,wec an us ea c l as s i c als pac i ngf i l l i ng s ampl i ng t e c hni 6ues uc h asLat i n ,(s ,af hype r c ubet os e l e c taf e wt r i alpoi nt s e e445).I nt hi spape r i ni t es e tofpoi nt si s obt ai ne dbyapr oc e s ss i mi l art oge ne t i cal gor i t hm i ns e ar c hs t e p.Thi si de ai ss t i mul at e dby 455,butt ,t ,whe he yar edi f f e r e nti nnat ur e.Fi r s t heal gor i t hm i n455i spr obabi l i s t i c r e as t heal gor i t hm i nt hi spape ri sade t e r mi ni s t i cpat t e r ns e ar c hal gor i t hm.7nt heot he rhand, .8ot onl t he c onve r ge nc e anal ys i si s di f f e r e nt y i st he al gor i t hm de t e r mi ni s t i c al l y 0321. Re c e i ve d:2005MR Subj e c tCl a s s i f i c a t i on:90C30,90C56. :unc . Ke ywor ds ons t r a i ne dopt i mi z a t i on,pa t t e r ns e a r c hme t hod,ge ne t i ca l gor i t hm,gl oba lmi ni mi z e r Suppor t e dbySc i e nt i f i cRe s e a r c h FundofHuna nPr ovi nc eEduc a t i onCommi t t e e(04C464)a ndby Hua i huaCol l e ge. .yat .B AXXl h.J.Che [e s eU[e ‘.Se r 49W .Wl,No.4 Vol ,butal c onve r ge nt s oi t snume r i c alr e s ul t ss how i tc anf i ndt hegl obalmi ni mi z e rofpr obl e m (1),whi c hi ss i mi l art ot hege ne t i cal gor i t hm. ;< =>?@ A BCD EF@ A G HI Ji r s twei nt r oduc eade f i ni t i oni nK1LM [ T UV1,VW,...,VXYbeas NO P A QA G A FQ.Re tS e tofXZ [\ 1ve c t or si n] .^he nt hes e tS [ [ ...,c ge ne r at e s] ,i ff oran_ve c t or‘a] ,t he r ee bi s tnonne gat i ve sc c uc ht hat‘T 1, W, Xs X d cV. e e e T1 [ [ j TUV1,VW,...,VXYge T1,W,...,X),t ne r at e] .h fia] andi VkZl,(k he n fO IICg.Re tS iTl. mHO=>?@ A BnD EF@ A G HIM oG O pq ri ve nt hei ni t i als t e ps l e ngt hc ont r olpar ame t e rt1ul,t hec ont r ac t i ngf ac t orof s t e ps l e ngt hc ont r olpar ame t e rlv wv 1 and t he above bound ofs t e ps l e ngt hc ont r ol xy .Re par ame t e rtmabTw 1t1,whe r ey1u1i sapos i t i vei nt e ge r tlvttolz1bet het ol e r anc e us e dt ot e s tf orc onve r ge nc e.ri ve nt henumbe rofi ndi vi dualofe ve r _ge ne r at i on{,andt he pos s i bi l i t _ ofc r os s ove rand mut at i on X|and X}.ri ve n al s o age ne r at i ng s e tS hat 1 t [ ge ne r at e s] .re ne r at et hef i r s tge ne r at i on~1TUx1,1,x1,W,...,x1,{Y, T1,W,...,{Y.k:T1. gmi nUf(x1,e),e x1Tar oG O pg Pe r f or mt hee vol ut i onar _pr oc e s st oge t~k\ 1TUxk\ 1,1,xk\ 1,W,...,xk\ 1,{Y; T1,W,...,{Y,i oG O p< x Tar gmi nUf(x ),e ff(x )vf(x),t he ns e t k\ 1 k\ 1,e k\ 1 k - , Δk\ 1 T mi nUΔmab,w ΔkY,xk\ 1 T x k\ 1 x1 andgot oSt e p5; he ns e txk\ 1 T xk \ ΔkVk; oG O p3 h f∃Vka S uc ht hatf(xk\ΔkVk)vf(xk),t ks oG O p4 Ot he r wi s ef ore ve r _Va S f(xk\ΔkV)Zf(xk).h nt hi sc as es e t k, xk\ 1 T xk andΔk\ 1 T wΔk; i fΔk\ 1vΔtol,t he nt e r mi nat e; oG O p5 Updat et hege ne r at i ngs e tS oS k:Tk\1,got oSt e p1. kt k\ 1. M mHOO vFD uG A FQC@ >p@ Fc O s s oG O pg.g Sor tt hei ndi vi dual si n~kac c or di ngt ot hef unc t i onval uei nde s c e ndi ngor de r t oobt ai n~kTUxk,1,xk,W,...,xk,{Y; -M oG O pg.< B_t hef ol l owi ngwa_c r os s ove rt hef i r s ty(1vyv{)i ndi vi dual si n~ k T1t Jore oy do h fr df()vp| ({); e di nt 1Tr ({); e di nt WTr ([); e di nt 3Tr A DE3I 4A{I 4E5 F3EEA6783I {9M F83 ):;8:b{3AI :ED 8<{I MI =A{I 8: . ,s +s n,-s ./i 012 4*3 Di s pl ac et hef i r s ti ompone nt sofve c t orxk,i2 byt hef i r s ti 3c 3 c ompone nt sofve c t orx t oge tane wi ndi vi dualx , k,i 1 k+ 1,i T .e.,],...,i xk+ 1,i:=(xk,i1[1],...,xk,i1[i xk,i2[i xk,i2[n]), 3] 3+1 - ]i whe r exk,j[i st hei t hc ompone ntofve c t orxk,j. El s e x - ; =x k,i k+ 1,i Endi f Endf or =M+1t Fori oN do - =x ; x k+ 1,i k,i Endf or St e p1.3 Mut at i on: =1t Fori oN do I fr di nt f()<pm (|^ ); i di nt 4=r k| xk+ 1,i=xk+ 1,i+_k‘i4; El s e xk+ 1,i=xk+ 1,i; Endi f Endf or ,f I nt hepr oc e s s unc t i onr di nt f()pr oduc e sar andom val uei n [a,1]andt hef unc t i on (N)pr r di nt oduc e sar andom i nt e ge rnumbe rbe t we e n1andN. bt e p1c anbel ooce dast hes e ar c hs t e pi nt hege ne r alpat t e r ns e ar c hme t hod. d3 efgh ij k jlmt neo lfpe q re fo e ,l Forc onve ni e nc e,i nt hi spape r e ts e not et hei t hc oor di nat edi r e c t i onands e tt= id uk: f(xk+ 1)< f(xk)v,w= uk: f(xk+ 1)= f(xk)v.I fkx t,wec al lt hekt hi t e r at ei s . s uc c e s s f ul : I nor de rt oge tt hegl obalc onve r ge nc ewene e dt hef ol l owi ngc ommonas s umpt i ons } = | ej j yzpt k lf1.{hes e t^ k= 1 ^isafiniteset.buppose^= u‘,‘,...,‘ vandallthe k 1 2 p .{heas ve c t or si n^ar er at i onal s umpt i onc anbes at i s f i e de as i l y,e.g.ate ve r yi t e r at el e t ^=u~s,~s,...,~sv. k 1 2 n . lve c t or si nx1 ar er at i onal ej j yzpt k lf2.Al ρ1 + ,whe r eρ1,ρ2xZ . ρ2 ej j yzpt k lf4.{hel e ve ls e t’f(x1)=ux|f(x)(f(x1)vi sbounde d. ej j yzpt k lf3.ρi sar at i onalnumbe randρ= . )nde rt he s eas s umpt i onst hef ol l owi ngl e mmasar eobvi ous .B .Mat ne s eUni v.Se r Appl h.J.Chi 494 .20,No.4 Vol T n ...,v Le mma2.For∀j∈ {1,2,...,p},t he r ee xi s tv v v ndapos i t i ve j= ( j ,1, j ,2, j ,n) ∈ Z a n r at i onalnumbe rbs uc ht hatdj = b Σ vj,iei. i =1 T n ∈{1,2,...,N},t Le mma3.Foral lj he r ee xi s tµj=(µj,1,µj,2,...,µj,n)∈Z andapos i t i ve n e . r at i onalnumbe rb uc ht hatx1,j = b xs xΣ µ j ,i i i =1 Г(k) Г(k) - Г(k) Le mma4.Δk=ρ Δ1=ρ1 ρ2 Δ1,whe r eГ(k)∈Zi sbounde dbe l ow wi t h-M. Now we s t r i ve t o ge tt he i mpor t ant c onc l us i on i nc onve r ge nc e anal ys i s ofpat t e r n .Fors al gor i t hms i mpl i c i t y,l e txk,0 de not exk. T n Le mma5.Foral lk=1,2,...,t he r ee xi s tωk,j= (ωk,j,1,ωk,j,2,...,ωk,j,n) ∈Z ,βk,j= (βk,j,1, T n βk,j,2,...,βk,j,n)∈Z s uc ht hat n n Г1(k) Г2(k) xk,j = b e ρ1 ρ2 Δ1Σ βk,j,ie 0,1,2,...,N}, xΣ ω k,j ,i i+ b i j∈ { i =1 (2) i =1 andf ore ve r yωk,j,it he r ee xi s tj i uc ht hatωk,j,i=µj0,i0,whe r eГ1(k),Г2(k)∈Z. 0, 0s .Thepr Pr oof oofi sbyi nduc t i on.Fork= 1,Le mma3i mpl i e st hec onc l us i on.Now as s ume (2)hol ∈{0,1,...,N}.The . dsf orxk,j,j r ear ef ourpos s i bl eout c ome s 1.xk+ 1,j=xk,j.I nt hi sc as e,wewr i t e n n i =1 i =1 Г(k) Г2(k) xk+ 1,j = b e ρ1 ρ2 Δ1Σ βk+ 1,j,ie , xΣ ω k+ 1,j ,i i+ b i (3) whe r eωk+ 1,j,i=ωk,j,iandβk+ 1,j,i=βk,j,i∈Z. 2.xk+ 1,ji spr oduc e dbyc r os s i ngxk,j1 andxk,j2 ati nde xi I nt hi sc as e, 0. i 0 i 0 n xk+ 1,j =b x( Σ ωk,j1,iei+ i =1 Σω e)+ b ρ k,j ii 2, i =i 0+ 1 n Г1(k) Г2(k) 1 2 1 n n i =1 i =1 ρ Δ (Σ βk,j1,ie i+ i =1 Σβ e)= k,j ii 2, i =i 0+ 1 Г1(k) Г2(k) b e ρ1 ρ2 Δ1Σ βk+ 1,j,ie ; xΣ ω k+ 1,j ,i i+ b i (4) ≤i whe r eωk+ 1,j,i= ωk,j1,i,βk+ 1,j,i= βk,j1,i∈Z,(1≤i ndωk+ 1,j,i= ωk,j2,i,βk+ 1,j,i= βk,j2,i∈ Z, 0)a (i ≤n). 0<i - whi 3.Wege tt hepoi ntxk+ 1,jbyaddi ngΔkdj3 t ox c hi spr oduc e dbyc r os s i ngxk,j1 and k,j xk,j2 ati nde xi Fort hi sc as e, 0. i 0 n xk+ 1,j =b x( Σ ωk,j1,iei+ i =1 Σω e)+ k,j ii 2, i =i 0+ 1 i 0 n Г1(k) Г2(k) 1 2 1 b ρ ρ Δ (Σ βk,j1,ie i+ i =1 n Г(k) - Г(k) e)+ b ρ1 ρ2 Δ1Σ v e j i i= 3, Σβ k,j ii 2, i =i 0+ 1 n n i =1 i =1 i =1 ' (k) Г' 1 b e ρГ ρ22(k)Δ1Σ βk+ 1,j,ie ; xΣ ω k+ 1,j ,i i+ b 1 i (5) (k)= mi (k)= mi ≤i whe r eГ1' n{Г1(k),Г(k)},Г2' n{Г2(k),- Г(k)}.I f1≤ i ωk+ 1,j,i= 0, Г1(k)- Г' (k) Г2(k)- Г' (k) Г(k)- Г' (k) - Г(k)- Г' (k) 1 2 1 2 =ρ ≤ n,ω = ω ,β ρ β +ρ ρ v ∈ Z.I fi< i k,j i 1, k+ 1,j ,i 1 2 k,j i 1, 1 2 j i 3, 0 k+ 1,j ,i _ 234I 5_xI 536 Z433_a784I x9: Z84 n;’8;Sx4_I ;32 8<xI :I =_xI 8; . ,Q *Q P+,Q -.i /01 4lb Г (k)- Г' (k) Г2(k)- Г' (k) Г(k)- Г' (k) - Г(k)- Г' (k) 1 2 1 2 ∈Z. ωk,j2,i,βk+ 1,j,i=ρ11 ρ2 βk,j2,i+ρ1 ρ2 v j i 3, 4.xk+ 1,0 i sobt ai ne dbySt e p3,asar e s ul tt he r ee xi s t sdj3∈ L uM ht hat xs P P P Г1(k) Г2(k) 1 1 2 xk+ 1,0 =N Q ρ xO ω k,0,i i+ N Г(k) - Г(k) 1 1 2 R O βk,0,iQ ρ ρ i+ N ρ i =1 RO v Q j i i= 3, i =1 P P i =1 i =1 i =1 S k) ГS 1( Q ρГ ρ2 2(k)R1O βk+ 1,0,iQ T N xO ω k+ 1,0,i i+ N i 1 (U) Vhe r eГS k)=Wi nXГ1(k),Г(k)Y,ГS k)=Wi nXГ2(k),-Г(k)Y∈Z,ωk+ 1,0,i=ωk,0,i,βk+ 1,0,i 1( 2( Г1(k)- ГS k) Г2(k)- ГS k) Г(k)- ГS k) - Г(k)- ГS k) 1( 2( 1( 2( =ρ1 ∈Z. ρ2 βk,0,i+ρ1 ρ2 v Q j i i 3, Zr oW (3)[(U),l e t Г1(k+ 1)= Wi nXГ1(k),Г1' (k),Г1S (k)Y= Wi nXГ1(k),Г(k)Y∈ Z, (\) nXГ2(k),Г2' (k),Г2S (k)Y,= Wi nXГ2(k),- Г(k)Y∈ Z, Г2(k+ 1)= Wi (]) ^ P . t he r ee xi s t(βk+ 1,j,1,βk+ 1,j,2,...,βk+ 1,j,P)∈Z s uM ht hat(2)hol ds . _sadi r e M ti n‘ e r e nM eo‘ae WWab,t he‘ ol l oVi ncl e WWai sobdi ousandi Wpor t ant ^ P ...,j ..., ef gghi.Zoral lk=1,2,...,t he r ee xi s tj j j k k k k=( k,1, k,2, k,P)∈Z , k=( k,1, k,2, ^ P k uM ht hat k,P)∈Z s P P Г1(k) Г2(k) Q ρ1 ρ2 R1O k Q , xk = N xO j k,i i+ N k,i i i =1 (l) i =1 =mj0,i0,Vhe and‘ ore de r yj he r ee xi s tj i uM ht hatj r eГ1(k),Г2(k)∈Z,andГ1(k)= k,it 0, 0s k,i Wi nXГ1(k-1),Г(k-1)Y,Г2(k)=Wi nXГ2(k-1),-Г(k-1)Y. nnde r_s s uWpt i ons1[4,ae WWaUl e adst ot he‘ ol l oVi nct he or e W Vhi M hi sne M e s s ar y . ‘ oral lpat t e r ns e ar M hal cor i t hWs . opf qr f g s.l i Wi n‘ kt u R k=0 .Suppos .xhe vr qqw enot nt he r ee xi s t sRy z0s uM ht hatRkzRy ‘ oral lk.xhi sboundonRk ,byae i nt ur ni Wpl i e s WWa4andae WWaU,t hatt he r eWus te xi s taM ons t antГy s uM ht hat oral lk.xhe n(l)M anber e Vr i t t e nas Г(k){Гy ,Г1(k){Гy andГ2(k){Гy ‘ P P i =1 i =1 | y y Гy xk = N Q ρГ Q , xO j k,i i+ N 1 ρ 2 R 1O k k,i i (10) y k)- Гy Г2(k)- Гy | 1( =ρГ ∈Z. Vhe r ek ρ2 k k,i 1 k,i N N ' Гy Гy 1 + ,N ' ,N S ,N ' ,N ρ1 ρ2 R0= ,Vhe r eN N ae t(N e not et hecr e at e s t 1, 2∈ Z . 1)d N S N 2 'andN S ,N SandN di di s orM oWWono‘N ae t}N e not et hel e s sM oWWonWul t i pl eo‘N xhe n 1. 2~d 2. ae tN x= (10)M anber e Vr i t t e nas P (N ' ,N N ' }N S ,N N N S ,N 1) 2~ 1} 2~ | y y y xk = k Q ,k j k . k,i k,i= k,i∈ Z i k,i+ O }N S ,N N S ( N ' , N ) N ( N ' , N ) 2~i 1 2 1 =1 (11) (11)I l l us t r at e st he‘ aM tt hat‘ oral lkt hei t e r at epoi ntxk Wus tl i eonal at t i M e" (N ' ,N 1) z0Vhos . s M al e dbyaM ons t ant edi r e M t i onsar epar al l e lt ot heM oor di nat edi r e M t i ons }N S ,N 2~ Si nM eanys uM M e s s ‘ uli t e r at eWus tbei nt he‘ i ni t es e t"# $%(x1),andnoi t e r at eM an bes uM M e s s ‘ ulWor et han onM e, i t‘ ol l oVst hat Wus tbe‘ i ni t e.’ons e (ue nt l y i s .{at .B Avvl h.J.Chi n~ s ~Uni v.S~ r 4xq .X0,No.4 nol i nf i ni t e.Ac c or di ngt ot heme c hani s m oft heal gor i t hm,Δke xpandsonl yf i ni t e l ymanyt i me s i mΔk = 0. andc ont r ac t si nf i ni t e l ymanyt i me sbyaf ac t orρ,whi c hguar ant e e st hatl k→ ∞ . Thi si sac ont r adi c t i onandt hec l ai mf ol l ows Unde rAs s umpt i ons1-4,The or e m 1l e adst ot hef ol l owi ng c onve r ge nc et he or y, whi c hi ss i mi l art ot heThe or e m 3.11i n[1]. Δ Δ W f(xk)W=0. he nl i mi nf sc ont i nuousi nt hel e ve ls e tVf(x1),t The or e m 2.I f f(x)i k→ ∞ The or e m Xs howst hef ac tt hatt he r ee xi s t satl e as tonel i mi tpoi ntofs e Yue nc eZxk[ t hati st hes t at i onar ypoi ntofpr obl e m (1). \e m]r ^._e e nf r om t heanal ys i sofc onve r ge nc e,t heal gor i t hm mus tbec onve r ge nti ft he e vol ut i onar ypr oc e s sc r os s e sbye xc hangi ngt hec ompone nt soft wove c t or sandmut at e sby . addi ngΔk‘k t os omeve c t or s ab cdme r e f ]ge hie r e me jk l I nt hi ss e c t i ont hehybr i dal gor i t hm i si mpl e me nt e dt os omepr obl e msi nt hee nvi r onm me ntofnopp qr0. ,s tΔ1= 1,Δtol= 0r00001,ρ= 0ru, si r s t e tt hehybr i dal gor i t hm par ame t e r sasf ol l ows =0rx,vy=0r001,z=u0,{1=uand{=Xu. vw .sors _e c ond,c hoos et hege ne r at i ngs e t i mpl i c i t y,ate ve r yi t e r at et hege ne r at i ngs e t |=Z}~,}~,...,}~[. k 1 X n ,t Las t her e s ul t sobt ai ne dbyhybr i dal gor i t hm ar el i s t e di nt hef ol l owi ngTabl e1, .de . whe r ei t e r not e st henumbe rofi t e r at e s Ta bl e1 Nume r i c a lRe s ul t soft hehybr i da l gor i t hm pr o mi ni mi z e r mi ni mum . i t e r 1 [0.000000,-1.000000] 3.000000 uq X [-1.4Xu1Xu,-0.8003X3] -18q.730x0x 37 3 [-0.08x844,0.71XqqX] -1.031qX8 34 4 [0.000000,0.000000] -X.000000 Xq u [x.000000,0.x1uuX7,-q.48437u,-X.3u3u1q,1.000000,1.000000] 0.000000 30 q [x.000000,0.x1uuX7,-q.48437u,-X.3u3u1q,1.000000,1.000000] 0.000000 30 7 [-x.u000000,-3.47x004,8.03100q,4.084473,1.000000,1.000000] 0.000000 48 8 [u.3qXX44,u.3qXX44,u.3qXX44,u.3qXX44,u.3qXX44, u.3qXX44,u.3qXX44,u.3qXX44,u.3qXX44,u.3qXX44] -1X.1ux8XX ux Att hes amet i me,t hege ne r alpat t e r ns e ar c hal gor i t hm i n[1]andt hege ne r alge ne t i c .The al gor i t hm i n [q] ar ec ode di n nopp q.0 and t e s t e d on t he s ame pr obl e ms A DERI VATI VEFREEALGORI THM FOR UNCONSTRAI NED OPTI MI ZATI ON ,e . Pe ngYe hui tal 497 c or r e s pondi ngpar ame t e r sus e dar et hes amewi t ht hos ei nt hehybr i dal gor i t hm.Thei ni t i al poi nt sar et hos ewhos ee ve r yc ompone nti s1.Thenume r i c alr e s ul t sar er e por t e di nt he Tabl e2andTabl e3r e s pe c t i ve l y. Ta bl e2 Nume r i c a lRe s ul t soft hege ne r a lpa t t e r ns e a r c ha l gor i t hm pr o mi ni mum i t e r 1 2 3 4 5 6 7 8 3.0000 -150.007436 -1.025494 -2.00000.000053 0.000043 0.000032 -10.300548 576 875 123 420 1126 1033 1204 2103 Ta bl e3 Nume r i c a lRe s ul t soft hege ne r a lge ne t i ca l gor i t hm pr o 1 2 3 4 5 6 7 8 . i t e r 57 38 30 27 40 40 45 82 Fr om t he s enume r i c alr e s ul t swec an e as i l yc omet oac onc l us i on t hatt hehybr i d al gor i t hm e xc e l st hege ne r alpat t e r ns e ar c h al gor i t hm and ge ne t i cal gor i t hm i ns ome . as pe c t si ns omec as e s Te s t i ngf uc nt i ons (1)Gol ds t e i nPr i c ef unc t i on: f(x1,x2)=[1+ (x1 + x2 + 1)2(19- 14x1 + 3x2 4x2 + 6x1x2 + 3x2 ]· 1- 1 2) [30+ (2x1 - 3x2)2(18- 32x1 + 12x2 8x2 - 36x1x2 + 27x2 ]. 1+ 4 2) =1,2). I thasonl yonegl obalmi ni mi z e rf(0,-1)=3i nr e gi on-2≤xi≤2(i (2)Shube r tf unc t i on: 5 f(x1,x2)= 5 Σ icos[(i+ 1)x1 + i]·Σ icos[(i+ 1)x2 + i]. i =1 i =1 = 1,2),t I nr e gi on- 10≤ xi≤ 10(i hi sf unc t i onhas760l oc almi ni mi z e r sandt hegl obal mi ni mum oft hi sf unc t i oni s-186.731. (3)Si xhumpCame lBac kFunc t i on: f(x,y)= (4- 2.1x2 + x4 2 )x + xy+ (- 4+ 4y2)y2. 3 Thi sf unc t i onhass i xl oc almi ni mi z e r si nr e gi on-3≤x≤3,-2≤y≤2,andt hegl obal mi ni mum i sf(-0.0898,0.7126)=f(0.0898,-0.7126)=-1.031628; (4)Ras t r i gi nf unc t i on: 2 f(x1,x2)= x2 os (18x1)- c os (18x2). 1+ x 2- c .Thegl I nr e gi on- 1≤ x1,x2≤ 1t hef unc t i onhasabout50mi ni mi z e r s obalmi ni mum i s f(0,0)=-2; (5)Si nes quar eⅠ f unc t i on(n=6): n- 1 f(x)= {10s i n2(πx1)+ (xn - 1)2 + 2 2 Σ (x - 1)[1+ 10sin(πx i )]}π/n. i +1 i =1 = 1,2,...,6)t .Thegl I nr e gi on- 10≤xi≤ 10(i hi sf unc t i onhasabout60mi ni mi z e r s obal .Mat .B Appl h.J.Chi ne s eUni v.Se r 498 .20,No.4 Vol mi ni mum i sf=0; (6)Si nes quar eⅡ f unc t i on(n=6): n- 1 f(x)= {10s i n2(πy1)+ (yn - 1)2 + 2 2 Σ (y - 1)[1+ 10sin(πy i )]}π/n. i +1 i =1 yi= 1+ (xi- 1)/4 - 10≤ xi≤ 10. = 1,2,...,6)t .Thegl hi sf unc t i onhasabout30mi ni mi z e r s obal I nr e gi on- 10≤xi≤ 10(i mi ni mum i sf=0; (7)Si nes quar eⅢ f unc t i on(n=6): f(x)={10s i n2(3πx1)+ (xn - 1)2[1+ s i n2(2πxn)]+ n- 1 2 2 Σ (x - 1)[1+ sin(3πx i )]}/10. i +1 i =1 =1,2,...,6)t .Thegl I nr e gi on-10≤xi≤10(i hi sf unc t i onhasabout180mi ni mi z e r s obal mi ni mum i sf=0; Re f e r e nc e s 1 Kol daT G,Le wi sR M,Tor c z onV.Opt i mi z a t i onbydi r e c ts e a r c h:ne w pe r s pe c t i veons omec l a s s i c a l ,SI 482. a ndmode r nme t hods AM Re vi e w,2003,45(3):385- 2 Le wi sR M,Tor c z onV.Pa t t e r ns e a r c ha l gor i t hmsf orboundc ons t r a i ne dmi ni mi z a t i on,SI AM J our na l 1099. onOpt i mi z a t i on,1999,9:1082- 3 Aude t C,De nni s J E.A Pa t t e r n Se a r c h Fi l t e r Me t hod f or Nonl i ne a r Pr ogr a mmi ng wi t hout ,Te 09,De ,Ri De r i va t i ve s c hni queRe por t00pa r t me ntofComput a t i ona la nd Appl i e d Ma t hma t i c s c e Uni ve r s i t y,Hous t on,TX,2000. 4 St , e i nM.La r ges a mpl epr ope r t i e sofs i mul a t i onsus i ngLa t i n hype r c ubes a mpl i ng,Te c hnome t r i c s 1987,29:143151. 5 Ha r tW E.A c onve r ge nc ea na l ys i sofunc ons t r a i ne da nd bound c ons t r a i ne de vol ut i ona r y pa t t e r n 23. s e a r c h,Evol ut i ona r yComput a t i on,2001,9(1):1- 6 Abdul l a hA R.A r obus tme t hodf orl i ne a ra ndnonl i ne a ropt i mi z a t i onba s e donge ne t i ca l gor i t hm, 287. Cybe r ne t i c a,1991,34:279- 1 Sc .a .Te hoolofMa t h.Sc i ndComput c h.,Ce nt r a lSout hUni v.,Cha ngs ha410075,Chi na. 2 De .ofMa pt t h.,Hua i huaCol l e ge,Huna n418008,Chi na.