()82JOURNALOFGUANGXIUNIVERSITYFORNATIONALITIESVol.8No.22002年5月(NaturalScienceEdition)May.2002:1007-0311(2002)02-0013-02X凌征球(玉林师范学院数计系,广西玉林537000):、,、.,.:;;:O17411:A1设n元数值函数y=f(x1,x2,,,xn)在x0=(x01,x02,,,x0n)点可偏导,则称向量:f(x0)x1,f(x0)x2,,,f(x0)xn为函数y=f(x1,x2,,,xn)在点x0的梯度,记作gradf(x0),即gradf(x0)=f(x0)x1,f(x0)x2,,,f(x0)xn2设函数y=f(x1,x2,,,xn)在x0=(x01,x02,,,x0n)点有连续的二阶偏导数,称矩阵Hf(x0)=fMx1x1(x0)fMx1x2(x0),fMx1xn(x0)fMx2x1(x0)fMx2x2(x0),fMx2xn(x0),,,,fMxnx1(x0)fMxnx2(x0),fMxnxn(x0)为函数y=f(x1,x2,,,xn)在x0点的海色矩阵.1如果点x0=(x01,x02,,,x0n)是函数y=f(x1,x2,,,xn)的极值点,并且存在偏导数fxi(i=1,,,n),则gradf(x0)=0证明:如果点x0=(x01,x02,,,x0n)是函数y=f(x1,x2,,,xn)的极值点,则一元函数f(x1,x02,,,x0n)在点x01有极值,于是由费尔马定理,f(x1,x02,,,x0n)x1x1=x01=0,同理,f(x01,x2,,,x0n)x2x2=x02=0,,,,f(x01,x02,,,xn)xnxn=x0n=0因此,gradf(x0)=f(x0)x1,f(x0)x2,,,f(x0)xn=0.2设函数y=f(x1,x2,,,xn)在x0的某个邻域内有连续的二阶偏导数,如果y=f(x1,x2,,,xn)在点x0处取极小值(极大值),则矩阵Hf(x0)是正定或半正定矩阵(负定或半负定矩阵)证明:作函数g(t)=f(x0+tüx),üx=(üx1,,,üxn),则g(t)在t=0时取极小值,因此,gd(0)0不可能成立,否则在t=0时取极大值,于是gd(0)E0.但gd(0)=(üx1,,,üxn)Hf(x0)üx1süxn,即对任意的üx=(üx1,,,üxn)都有gd(0)=(üx1,,,üxn)Hf(x0)üx1süxnE0,13X:2001211215.:凌征球(19632),广西桂平人,玉林师范学院数计系讲师.所以Hf(x0)是正定或半正定矩阵.3设函数y=f(x1,x2,,,xn)在x0点处有连续的二阶偏导数,并且gradf(x0)=0,则(1)当矩阵Hf(x0)是正定矩阵时,y=f(x1,x2,,,xn)在x0处取得极小值;(2)当矩阵Hf(x0)是负定矩阵时,y=f(x1,x2,,,xn)在x0处取得极大值.证明:令x=(x1,x2,,,xn),x0=(x01,x02,,,x0n)考虑函数y=f(x1,x2,,,xn)在x0点的Taylor展开公式:f(x)=f(x0)+Eni=1f(x0)xi(xi-x0i)+12Eni,j=12f(x0)xixj(xi-x0i)(xj-x0j)+0(|x-x0|2)=f(x0)+f(x0)x1,,,f(x0)xnx1-x01sxn-x0n+12(x1-x01,,,xn-x0n)fMx1x1(x0)fMx1x2(x0),fMx1xn(x0)fMx2x1(x0)fMx2x2(x0),fMx2xn(x0),,,,fMxnx1(x0)fMxnx2(x0),fMxnxn(x0)x1-x01x2-x02sxn-x0n+0(|x-x0|2)=f(x0)+gradf(x0)#üx+12üxc#Hf(x0)#üx+0(|üx|2)因为gradf(x0)=0,所以,f(x)-f(x0)=12üxc#Hf(x0)#üx+0(|üx|2)因此,如果矩阵Hf(x0)二次型üxc#Hf(x0)#üx是正定二次型,üxc#Hf(x0)#üx0于是,在|üx|足够小时,有f(x)-f(x0)0,即f(x)在x0处取极小值.同样,如果矩阵Hf(x0)是负定矩阵时,二次型üxc#Hf(x0)#üx是负定二次型,即:üxc#Hf(x0)#üx0,在|üx|足够小时,有f(x)-f(x0)0,f(x)在x0处取极大值.1设n元函数y=f(x1,x2,,,xn)在x0=(x01,x02,,,x0n)的某个邻域内有连续的二阶偏导数,且gradf(x0)=0,则(1)若di(x0)0,(i=1,2,,,n)则x0为f(x)的极小值点;(2)若(-1)idi(x0)0,(i=1,2,,,n)则x0为f(x)的极大值点.其中di(x0)=fMx1x1(x0),fMx1xi(x0),,,fMxix1(x0),fMxixi(x0)事实上,di(x0)就是海色矩阵Hf(x0)的所有主子式,如果所有的主子式都大于零,则矩阵Hf(x0)是正定矩阵.2对于一元函数y=f(x),设f(x)在x0处有二阶连续偏导数,且fc(x0)=0,则在fd(x0)0时,f(x)在x0处取极小值,在fd(x0)0时,f(x)在x0处取极大值.事实上,海色矩阵Hf(x0)=(fd(x0)),若Hf(x0)是正定矩阵,那么行列式|Hf(x0)|=fd(x0)0,因此,f(x)在x0取极小值,同样,Hf(x0)是负定矩阵时,行列式|Hf(x0)|=fd(x0)0,f(x)在x0取极大值.3对于二元函数y=f(x1,x2),如果在x0=(x01,x02)处有二阶连续偏导数,且fcx1(x0)=fcx2(x0)=0,fdx1x1(x0)#fdx2x2(x0)-(fdx1x2(x0))20,则在fdx1x1(x0)0时,函数y=f(x1,x2)在x0点处取极小值,在fdx1x1(x0)0时,y=f(x1,x2)在x0点处取极大值.事实上,海色矩阵Hf(x0)=fdx1x1(x0)fdx1x2(x0)fdx2x1(x0)fdx2x2(x0),由已知,行列式|Hf(x0)|=fdx1x1(x0)#fdx2x2(x0)-(fdx1x2(x0))20,因此,在fdx1x1(x0)0时,海色矩阵是正定矩阵,在fdx1x1(x0)0时,海色矩阵是负定矩阵.例求函数u=x3+3xy2-15x-12y的极植.解:求解方程组ux=0,uy=0,即3x2+3y2-15=06xy-12=0得四个稳定点:(2,1),(-2,-1),(2,1),(-1,-2)进一步计算得2ux2=6x,2uyx=6y,2uy2=6x矩阵H(2,1)=126612是正定矩阵,(2,1)是极小值点.H(-2,-1)=-12-6-6-12是负定矩阵,(-2,-1)是极大值点.H(1,2)=612126,H(-1,-2)=-6-12-12-6均是不定矩阵,(1,2),(-1,-2)均不是极值点.[][1],1()[M].:,19991[2],1[M].:,20001[](28)14()200258在实际应用中只需测定资源种群生态学参数r和k,并把种群大小调节k/2到值以上,并保持rk/4的收获量,即可实现持续的最大产值,根据以上分析可以提出如下的管理对策:(1)当放牧或刈割使牧草种群xk/2时,可以加大放牧强度,使种群降到k/2值.(2)当放牧或刈割使种群xk/2时(即dx/dt0时,种群将趋于零,xv0,这将导致种群灭绝),这时应减少放牧强度(甚至关闭牧场),使x上升到k/2值以上.(3)当放牧或刈割使牧草种群x=k/2时,应维持这个放牧强度,这时的放牧强度即为适度的放牧强度.关于参数r的测定,可用如下方法估计:当一个种群侵入空白地的初期,它基本是按指数的规律增长,即x=x0ert(3)其中:x0为t=0时的种群大小,x=x(t)是经过t时刻后种群的大小,只要适当取t值,上述规律是对的,因为对(3)两边取对数得:r=1t(1nx-lnx0)(4)利用(4)式和连续测定的统计数据即可估计r值.对于环境容量k,可通过观察统计数据确定.[][1]1,[N].,1998-08-11,(7).[2]#[N].,2002-02.[3]1[J].,2001.[4]1[M]1:,2001.[3]1[J].(),2001,(5).[]TheStatisticalanalysisofForestcoverrateandSoilerasionandDesertificationLIUYou2lian(PreparatoryDept.ofGuangxiUniversityfornationalities,Nanning530006,China)Abstract:Onthebasisofthedependedrelationshipbetweenthethreevariables-forestcoverrate,soilerasionanddesertification,thisarticlefindsouttherequlationbetweenthembystatisticalycomparingmeans,andgivescorrespondlytheresolution1KeyWords:forest,coverrate,soilerasion,desertification(14)TheapplicationofthequadraticforminthefunctionofmanyvariableextremevalueLINGZheng2qiu(Dept.ofMathematics&ComputerScience,YulinTeachersCollege,Yulin537000,China)Abstract:Thecalculationsofextremevalueformonadicandbinaryfunctionareessentialprobleminmathematicalanalysis,whichcanapplytocalculatemaximumandminimum.Thispaperusesthetheoryofquadraticformtodistinguishthisproblem,notonlythat,itenlargetocalculateextremevalueforfunctionofmanyvariable.KeyWords:quadraticform,positivedefinitequadraticform,positivedefinitematrix,extremevalue,TaylorFor2mula28()200258