多元函数泰勒级数展开_多元函数的泰勒级数

一元函数的泰勒公式

equation?tex=f%28x%29+%3D+f%28a%29%2B%7B%5Cfrac+%7Bf%27%28a%29%7D%7B1%21%7D%7D%28x-a%29%2B%7B%5Cfrac+%7Bf%27%27%28a%29%7D%7B2%21%7D%7D%28x-a%29%5E%7B2%7D%2B%7B%5Cfrac+%7Bf%27%27%27%28a%29%7D%7B3%21%7D%7D%28x-a%29%5E%7B3%7D%2B%5Ccdots+%5Ctag%7B1%7D+

或者可以写成:

equation?tex=f%28x%29+%3D+%5Csum+_%7Bn%3D0%7D%5E%7B%5Cinfty+%7D%7B%5Cfrac+%7Bf%5E%7B%28n%29%7D%28a%29%7D%7Bn%21%7D%7D%28x-a%29%5E%7Bn%7D+%5Ctag%7B2%7D

二元函数的泰勒级数

对 二元函数 f(x, y), 考虑在点 (a, b) 附近方向 (u, v) 有微小增量 f(a + tu, b + tv), 定义函数:

equation?tex=%5Cphi%28t%29+%3A%3D+f%28a+%2B+tu%2C+b+%2B+tv%29+%5Cquad+%280+%5Cle+t+%5Cle+1%29%5C%5C

此时

equation?tex=%5Cphi%28t%29 是对 t 的单变量函数,利用上面的 一元函数的泰勒公式, 把

equation?tex=%5Cphi%28t%29 在 t = 0 处展开:

equation?tex=+%5Cphi%28t%29+%3D+%5Cphi%280%29+%2B+%5Cphi%27%280%29+t+%2B+%5Cfrac%7B%5Cphi%27%27%280%29%7D%7B2%21%7D+t%5E2+%2B+%5Ccdots+%5C+%5C%5C

取 t = 1:

equation?tex=%5Cphi%281%29+%3D+%5Cphi%280%29+%2B+%5Cphi%27%280%29+%2B+%5Cfrac%7B%5Cphi%27%27%280%29%7D%7B2%21%7D+%2B+%5Ccdots++%5C%5C

明显:

equation?tex=%5Cphi%28t%29+%3D+f%28a+%2B+tu%2C+b+%2Btv%29+%5C%5C

根据链式法则:

equation?tex=%5Cfrac%7Bdf%7D%7Bdt%7D+%3D+%5Cfrac%7B%5Cpartial+f%7D%7B%5Cpartial+x%7D+%5Cfrac%7Bdx%7D%7Bdt%7D+%2B++%5Cfrac%7B%5Cpartial+f%7D%7B%5Cpartial+y%7D+%5Cfrac%7Bdy%7D%7Bdt%7D+%5C%5C

有:

equation?tex=+%5Cbegin%7Balign%7D+%5Cphi%27%28t%29+%3D+%5Cfrac%7Bd%7D%7Bdt%7D%5Cphi%28t%29+%26%3D+%5Cfrac%7Bd%7D%7Bdt%7Df%28a+%2Btu%2C+b%2B+tv%29+%3D+f_x%28a+%2B+tu%2C+b+%2B+tv%29u+%2B+f_y%28a+%2Btu%2C+b%2Btv%29v+%5C+%5Cend%7Balign%7D+%5C%5C

所以

equation?tex=%5Cphi%27%280%29+%3D+f_x%28a%2C+b%29u+%2B+f_y%28a%2C+b%29v . 再次运用 链式法则 求导:

equation?tex=+%5Cbegin%7Balign%7D+%5Cphi%27%27%28t%29+%26%3D+%5Cfrac%7Bd%7D%7Bdt%7D%5Cbig%28f_x%28a+%2B+tu%2C+b+%2B+tv%29u+%2B+f_y%28a+%2Btu%2C+b%2Btv%29v%5Cbig%29+%5C%5C+%26%3D+f_%7Bxx%7D%28a%2Btu%2C+b%2Btv%29u%5E2+%2B+2f_%7Bxy%7D%28a%2Btu%2C+b%2Btv%29uv+%2B+f_%7Byy%7D%28a%2Btu%2C+b%2Btv%29v%5E2+%5C+%5Cend%7Balign%7D+

所以

equation?tex=%5Cphi%27%27%280%29+%3D+f_%7Bxx%7D%28a%2Cb%29u%5E2+%2B+2f_%7Bxy%7D%28a%2Cb%29uv+%2B+f_%7Byy%7D%28a%2Cb%29v%5E2 , 可以继续求得更高阶的导数,最终代回

equation?tex=+t+%3D+1+%5C%5C+u+%3D+x+-+a+%5C%5C+v+%3D+y+-+b++%5C%5C

可得二元函数在 (a,b) 的 泰勒级数展开:

equation?tex=%5Cbegin%7Balign%7D+f%28x%2Cy%29+%3D+%26f%28a%2Cb%29+%2B+f_x%28a%2Cb%29%28x+-+a%29+%2B+f_y%28y-b%29+%5C%5C+%26%2B+%5Cfrac%7B1%7D%7B2%21%7D%5Bf_%7Bxx%7D%28x-a%29%5E2+%2B+2f_%7Bxy%7D%28a%2Cb%29%28x-a%29%28y-b%29%2Bf_%7Byy%7D%28y-b%29%5E2%5D%2B+%5Ccdots+%5Cend%7Balign%7D

所以有时候我也看到这种写法:

equation?tex=%5Cbegin%7Balign%7D+f%28x_0+%2B+%5CDelta+x%2C+y_0+%2B+%5CDelta+y%29+%26%3D+f%28x_0%2C+y_0%29+%2B+%28%5CDelta+x+%5Cfrac%7B%5Cpartial%7D+%7B%5Cpartial+x%7D+%2B+%5CDelta+y+%5Cfrac%7B%5Cpartial+%7D%7B%5Cpartial+y%7D%29+f%28x_0%2C+y_0%29+%5C%5C+%26%2B+%5Cfrac%7B1%7D%7B2%21%7D%28%5CDelta+x+%5Cfrac%7B%5Cpartial%7D+%7B%5Cpartial+x%7D+%2B+%5CDelta+y+%5Cfrac%7B%5Cpartial+%7D%7B%5Cpartial+y%7D%29%5E2+f%28x_0%2C+y_0%29+%2B+%5Ccdots+%5Cend%7Balign%7D+%5C%5C

第一次看到这种写法还真是让我疑惑了一会,o(╯□╰)o, 但是其实展开,是一样的:

equation?tex=%28%5CDelta+x+%5Cfrac%7B%5Cpartial%7D+%7B%5Cpartial+x%7D+%2B+%5CDelta+y+%5Cfrac%7B%5Cpartial+%7D%7B%5Cpartial+y%7D%29%5E2++%3D+%5CDelta+x+%5E2+%5Cfrac%7B%5Cpartial%5E2%7D+%7B%5Cpartial+x%5E2%7D++%2B+2+%5CDelta+x+%5Cfrac%7B%5Cpartial%7D+%7B%5Cpartial+x%7D%5CDelta+y+%5Cfrac%7B%5Cpartial+%7D%7B%5Cpartial+y%7D+%2B+%5CDelta+y%5E2+%5Cfrac%7B%5Cpartial%5E2+%7D%7B%5Cpartial+y%5E2%7D+%5C%5C+

或者写成:

equation?tex=f%28x%2C+y%29+%3D++%5Csum+_%7Bn%3D0%7D%5E%7B%5Cinfty+%7D+%5Cfrac%7B1%7D%7Bn%21%7D+%28%5CDelta+x+%5Cfrac%7B%5Cpartial%7D+%7B%5Cpartial+x%7D+%2B+%5CDelta+y+%5Cfrac%7B%5Cpartial+%7D%7B%5Cpartial+y%7D%29%5En+f%28x_0%2C+y_0%29%5C%5C

当然最让我亲切的写法还是跟 一元函数 类比/类似的写法。

一元函数线性近似

equation?tex=f%28x%29+%5Capprox+f%28a%29+%2B+f%27%28a%29%28x-a%29%5C%5C二次近似

equation?tex=f%28x%29+%5Capprox+f%28a%29+%2B+f%27%28a%29%28x-a%29+%2B+%5Cfrac%7Bf%27%27%28a%29%7D%7B2%21%7D%28x-a%29%5E2%5C%5C

二元函数线性近似

equation?tex=f%28x%2Cy%29+%5Capprox+f%28x_0%2Cy_0%29+%2B+f_x%28x_0%2Cy_0%29%28x-x_0%29+%2B+f_y%28x_0%2Cy_0%29%28y-y_0%29+%5C%5C

写成招人喜欢的向量形式:

equation?tex=f%28%5Cmathbf%7Bx%7D%29+%5Capprox+f%28%5Cmathbf%7Ba%7D%29+%2B+%5Cnabla+f%28%5Cmathbf%7Ba%7D%29+%5Ccdot+%28%5Cmathbf%7Bx%7D-%5Cmathbf%7Ba%7D%29%5C%5C二次近似

equation?tex=%5Cbegin%7Balign%7D+f%28x%2Cy%29++%26%5Capprox++f%28x_0%2Cy_0%29+%2B+f_x%28x_0%2Cy_0%29%28x-x_0%29+%2B+f_y%28x_0%2Cy_0%29%28y-y_0%29+%5C%5C+%26%2B+%5Cfrac%7B1%7D%7B2%7Df_%7Bxx%7D%28x_0%2Cy_0%29%28x+%E2%88%92+x_0+%29%5E2+%2B+f_%7Bxy%7D%28x-x_0%29%28y-y_0%29%2B+%5Cfrac%7B1%7D%7B2%7Df_%7Byy%7D%28x_0%2Cy_0%29%28y+%E2%88%92+y_0+%29%5E2%5C%5C+%5Cend%7Balign%7D+

同样写成向量形式:

equation?tex=f%28%5Cmathbf%7Bx%7D%29+%5Capprox+f%28%5Cmathbf%7Bx_0%7D%29+%2B+++%5Cnabla+f%28%5Cmathbf%7Bx_0%7D%29+%5Ccdot+%28%5Cmathbf%7Bx%7D+-+%5Cmathbf%7Bx_0%7D%29+%2B+%5Cfrac%7B1%7D%7B2%7D%28%5Cmathbf%7Bx%7D+-+%5Cmathbf%7Bx_0%7D%29%5ETH%28%5Cmathbf%7Bx_0%7D%29%28%5Cmathbf%7Bx%7D+-+%5Cmathbf%7Bx_0%7D%29%5C%5C

H 为 Hessian 矩阵:

equation?tex=H+%3D+%5Cbegin%7Bbmatrix%7D+%5Cfrac%7B%5Cpartial+f%7D%7B%5Cpartial+x%5E2%7D+%26++%5Cfrac%7B%5Cpartial+f%7D%7B%5Cpartial+x+%5Cpartial+y%7D+%5C%5C++%5Cfrac%7B%5Cpartial+f%7D%7B%5Cpartial+y+%5Cpartial+x%7D+%26++%5Cfrac%7B%5Cpartial+f%7D%7B%5Cpartial+y%5E2%7D+++%5Cend%7Bbmatrix%7D%5C%5C

可以做简单的验证:

equation?tex=%5Cbegin%7Balign%7D+%26%5Cbegin%7Bbmatrix%7D+x+-+x_0+%26+y+-+y_0+%5Cend%7Bbmatrix%7D+%5Cbegin%7Bbmatrix%7D+f_%7Bxx%7D+%26++f_%7Bxy%7D+%5C%5C++f_%7Byx%7D+%26+f_%7Byy%7D++%5Cend%7Bbmatrix%7D+%5Cbegin%7Bbmatrix%7D+x+-+x_0+%5C%5C+y+-+y_0+%5Cend%7Bbmatrix%7D+%5C%5C+%26%3D+%5Cbegin%7Bbmatrix%7D+%28x+-+x_0%29f_%7Bxx%7D+%2B+%28y+-+y_0%29f_%7Byx%7D+%26++%28x+-+x_0%29f_%7Bxy%7D+%2B+%28y+-+y_0%29f_%7Byy%7D+%5Cend%7Bbmatrix%7D+%5Cbegin%7Bbmatrix%7D+x+-+x_0+%5C%5C+y+-+y_0+%5Cend%7Bbmatrix%7D+%5C%5C+%26%3D+%28x+-+x_0%29%5E2f_%7Bxx%7D+%2B+%28y+-+y_0%29%28x-x_0%29f_%7Byx%7D+%2B++%28x+-+x_0%29%28y-y_0%29f_%7Bxy%7D+%2B+%28y+-+y_0%29%5E2f_%7Byy%7D+%5C%5C+%26%3D+%28x+-+x_0%29%5E2f_%7Bxx%7D+%2B++2%28x+-+x_0%29%28y-y_0%29f_%7Bxy%7D+%2B+%28y+-+y_0%29%5E2f_%7Byy%7D+%5Cend%7Balign%7D+%5C%5C

推广到变量更多的情况,写法也是多种多样,o(╯□╰)o:

比如写成这样:

equation?tex=%5Cbegin%7Balign%7D+f%28x_1%2C+x_2%2C+%5Ccdots%2C+x_m%29+%26%3D+f%28x_1%5E0%2C+x_2%5E0%2C+%5Ccdots%2C+x_m%5E0%29++%5C%5C+%26%2B+%5Cfrac%7B1%7D%7B1%21%7D+%28%5CDelta+x_1+%5Cfrac%7B%5Cpartial%7D+%7B%5Cpartial+x_1%7D+%2B+%5CDelta+x_2+%5Cfrac%7B%5Cpartial+%7D%7B%5Cpartial+x_2%7D+%2B+%5Ccdots+%2B+%5CDelta+x_m+%5Cfrac%7B%5Cpartial%7D%7B%5Cpartial+x_m%7D%29+f%28x_1%5E0%2C+x_2%5E0%2C+%5Ccdots%2C+x_m%5E0%29+%5C%5C+%26%2B+%5Cfrac%7B1%7D%7B2%21%7D%28%5CDelta+x_1+%5Cfrac%7B%5Cpartial%7D+%7B%5Cpartial+x_1%7D+%2B+%5CDelta+x_2+%5Cfrac%7B%5Cpartial+%7D%7B%5Cpartial+x_2%7D+%2B+%5Ccdots+%2B+%5CDelta+x_m+%5Cfrac%7B%5Cpartial%7D%7B%5Cpartial+x_m%7D%29+%5E2+f%28x_1%5E0%2C+x_2%5E0%2C+%5Ccdots%2C+x_m%5E0%29+%5C%5C+%26%2B+%5Ccdots+%5Cend%7Balign%7D%5C%5C

其中

equation?tex=%5CDelta+x_k+%3D+x_k+-+x_k%5E0 .

写成这样:

equation?tex=%5Cbegin%7Balign%7D+f%28x_1%2C%5Cldots%2Cx_d%29+%26%3D+f%28a_1%2C+%5Cldots%2Ca_d%29+%2B+%5Csum_%7Bj%3D1%7D%5Ed+%5Cfrac%7B%5Cpartial+f%28a_1%2C+%5Cldots%2Ca_d%29%7D%7B%5Cpartial+x_j%7D+%28x_j+-+a_j%29+%2B+%5Cfrac%7B1%7D%7B2%21%7D+%5Csum_%7Bj%3D1%7D%5Ed+%5Csum_%7Bk%3D1%7D%5Ed+%5Cfrac%7B%5Cpartial%5E2+f%28a_1%2C+%5Cldots%2Ca_d%29%7D%7B%5Cpartial+x_j+%5Cpartial+x_k%7D+%28x_j+-+a_j%29%28x_k+-+a_k%29+%5C%5C++%26+%5Cqquad+%5Cqquad+%2B+%5Cfrac%7B1%7D%7B3%21%7D+%5Csum_%7Bj%3D1%7D%5Ed%5Csum_%7Bk%3D1%7D%5Ed%5Csum_%7Bl%3D1%7D%5Ed+%5Cfrac%7B%5Cpartial%5E3+f%28a_1%2C+%5Cldots%2Ca_d%29%7D%7B%5Cpartial+x_j+%5Cpartial+x_k+%5Cpartial+x_l%7D+%28x_j+-+a_j%29%28x_k+-+a_k%29%28x_l+-+a_l%29+%2B+%5Ccdots+%5Cend%7Balign%7D%5C%5C

或者运用多重指标,写成上面的更加一元类似的形式:

equation?tex=f%28%5Cmathbf%7Bx%7D%29+%3D+%5Csum_%7B%7C%5Calpha%7C+%5Cgeq+0%7D%5Cfrac%7B%28%5Cmathbf%7Bx%7D-%5Cmathbf%7Ba%7D%29%5E%5Calpha%7D%7B%5Calpha+%21%7D+%5Cleft%28%7B%5Cmathrm%7B%5Cpartial%7D%5E%7B%5Calpha%7D%7Df%5Cright%29%28%5Cmathbf%7Ba%7D%29%5C%5C

推导更多元函数的泰勒级数 也可以用类似 二元函数 用 方向导数 的方式推出这个结论。

参考:托马斯微积分

wikipedia

评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值