线性变换和线性映射(二)
线性变换与线性映射在不同基与基偶下的关系
\(Thm1:\)若\(f\in Hom(V,U)\)在基偶\(V:α_{1},α_{2},...,α_{s},U:β_{1},β_{2},...,β_{n}\)下的矩阵是\(A\),\(\eta\in V\)在\(α_{1},α_{2},...,α_{s}\)下的坐标为\(X\),则\(f(\eta)\)在基\(β_{1},β_{2},...,β_{n}\)下的坐标为\(AX\)
\(Proof:\because(f(α_{1}),f(α_{2}),...,f(α_{s}))\)
\(=(β_{1},β_{2},...,β_{n})A\)
又\(\because\eta=(α_{1},α_{2},...,α_{s})X\),设\(X=(x_{1},x_{2},...,x_{s})^T\),
则\(\eta=x_{1}α_{1}+x_{2}α_{2}+...+x_{s}α_{s}\)
\(\therefore f(\eta)=f(x_{1}α_{1}+x_{2}α_{2}+...+x_{s}α_{s})\),又因为\(f\)为线性映射
有\(kf(x)=f(kx),f(x+y)=f(x)+f(y)\)
故\(\begin{equation}
\begin{aligned}
f(\eta)=&x_{1}f(α_{1})+x_{2}f(α_{2})+...+x_{s}f(α_{s})\\
=&(f(α_{1}),f(α_{2}),...,f(α_{s}))X
\end{aligned}
\end{equation}\)
即\(f(\eta)=(β_{1},β_{2},...,β_{n})AX\)
故\(f(\eta)\)在基\(β_{1},β_{2},...,β_{n}\)下的坐标为\(AX\)
\(Thm2:\)若\(f\in Hom(V,U)\)在基偶\(V:α_{1},α_{2},...,α_{s},U:β_{1},β_{2},...,β_{n}\)下的矩阵是\(A\),则\(f\)在新的基偶\((α_{1}^{'},α_{2}^{'},...,α_{s}^{'})=(α_{1},α_{2},...,α_{s})P\),\((β_{1}^{'},β_{2}^{'},...,β_{n}^{'})=(β_{1},β_{2},...,β_{n})Q\)下的矩阵\(B=Q^{-1}AP\)
\(Proof:\because(f(α_{1}),f(α_{2}),...,f(α_{s}))\)
\(=(β_{1},β_{2},...,β_{n})A\)
\(\begin{equation}
\begin{aligned}
&(α_{1}^{'},α_{2}^{'},...,α_{s}^{'})=(α_{1},α_{2},...,α_{s})P,\\
&(β_{1}^{'},β_{2}^{'},...,β_{n}^{'})=(β_{1},β_{2},...,β_{n})Q
\end{aligned}
\end{equation}\)
故\(f(α_{1}^{'},α_{2}^{'},...,α_{s}^{'})=f(α_{1},α_{2},...,α_{s})P\)
由于\(f\)为线性映射,故
\(\begin{equation}
\begin{aligned}
&f(α_{1}^{'},α_{2}^{'},...,α_{s}^{'})\\
=&(f(α_{1}^{'}),f(α_{2}^{'}),...,f(α_{s}^{'}))\\
=&(f(α_{1}),f(α_{2}),...,f(α_{s}))P
\end{aligned}
\end{equation}\)
故\(f(α_{1}^{'},α_{2}^{'},...,α_{s}^{'})=(β_{1},β_{2},...,β_{n})AP\)
又\((β_{1},β_{2},...,β_{n})=(β_{1}^{'},β_{2}^{'},...,β_{n}^{'})Q^{-1}\)
\(\therefore f(α_{1}^{'},α_{2}^{'},...,α_{s}^{'})=(β_{1}^{'},β_{2}^{'},...,β_{n}^{'})Q^{-1}AP\)
即\(f\)在新的基偶下的矩阵\(B=Q^{-1}AP\)
\(Thm3:\)若\(f\in Hom(V,V)\)在基\(α_{1},α_{2},...,α_{s}\)下的矩阵为\(A\),则\(f\)在新的基
\((α_{1}^{'},α_{2}^{'},...,α_{s}^{'})=(α_{1},α_{2},...,α_{s})P\)下的矩阵\(B=P^{-1}AP\)
\(Proof:\because(f(α_{1}),f(α_{2}),...,f(α_{s}))\)
\(=(α_{1},α_{2},...,α_{s})A\)
又\((α_{1}^{'},α_{2}^{'},...,α_{s}^{'})=(α_{1},α_{2},...,α_{s})P\)
故\(f(α_{1}^{'},α_{2}^{'},...,α_{s}^{'})=(f(α_{1}),f(α_{2}),...,f(α_{s}))P=(α_{1},α_{2},...,α_{s})AP\)
又\((α_{1},α_{2},...,α_{s})=(α_{1}^{'},α_{2}^{'},...,α_{s}^{'})P^{-1}\)
因此\(f(α_{1}^{'},α_{2}^{'},...,α_{s}^{'})=(α_{1}^{'},α_{2}^{'},...,α_{s}^{'})P^{-1}AP\)
即\(f\)在新的基下的矩阵\(B=P^{-1}AP\)
eg:求线性变换\(\begin{equation}
\begin{aligned}
&f:F_{3}[x]\longrightarrow F_{3}[x]:f(p(x))=p^{'}(x),\\
&\forall p(x)\in F_{3}[x]
\end{aligned}
\end{equation}\)在基
\(p_{1}(x)=1+x+3x^2,p_{2}(x)=1+x,\)
\(p_{3}(x)=1+2x-x^2\)下的矩阵
\(Sol1:\because \begin{equation}
\begin{aligned}
&p_{1}^{'}(x)=1+6x,p_{2}^{'}(x)=1,\\
&p_{3}^{'}(x)=2-2x
\end{aligned}
\end{equation}\)
设\(\begin{equation}
\begin{aligned}
&f(p_{1}(x),p_{2}(x),p_{3}(x))\\
=&(p_{1}^{'}(x),p_{2}^{'}(x),p_{3}^{'}(x))\\
=&(p_{1}(x),p_{2}(x),p_{3}(x))A
\end{aligned}
\end{equation}\)
解该线性方程得\(A=\begin{pmatrix}
\frac{5}{3} & -\frac{1}{3}& -\frac{4}{3}\\
-\frac{17}{3}& \frac{7}{3}& \frac{22}{3}\\
5& -1 &-4
\end{pmatrix}\)
\(Sol2:\)易知\(F_{3}[x]\)有一组基\({1,x,x^2}\),又\(\begin{equation}
\begin{aligned}
f(1,x,x^2)=&(1,x,x^2)\begin{pmatrix}
0& 1& 0\\
0& 0& 2\\
0& 0&0
\end{pmatrix}\\
=&(1,x,x^2)A
\end{aligned}
\end{equation}\),又
\(\begin{equation}
\begin{aligned}
&(p_{1}(x),p_{2}(x),p_{3}(x))\\
=&(1,x,x^2)\begin{pmatrix}
1& 1& 1\\
1& 1& 2\\
3& 0&-1
\end{pmatrix}\\=&(1,x,x^2)P
\end{aligned}
\end{equation}
\),
由\(Thm3\)可得本题所求矩阵为\(P^{-1}AP=\begin{pmatrix}
\frac{5}{3} & -\frac{1}{3}& -\frac{4}{3}\\
-\frac{17}{3}& \frac{7}{3}& \frac{22}{3}\\
5& -1 &-4
\end{pmatrix}\)
本人在学习这一部分知识时遇到的一些疑惑和自己的理解
- 在\(Thm3\)的证明中,对\((α_{1}^{'},α_{2}^{'},...,α_{s}^{'})=(α_{1},α_{2},...,α_{s})P\)两边施加线性变换\(f\)时,应为
\(f(α_{1}^{'},α_{2}^{'},...,α_{s}^{'})=f((α_{1},α_{2},...,α_{s})P)\),为何可以直接写成
\(\begin{equation} \begin{aligned} &f(α_{1}^{'},α_{2}^{'},...,α_{s}^{'})\\ =&f(α_{1},α_{2},...α_{s})P\\ =&(f(α_{1}),f(α_{2}),...,f(α_{s}))P\\ \end{aligned} \end{equation}\),即为何线性变换\(f\)对矩阵\(P\)没有作用
\(Sol:\)把\(P\)视为是一个数阵即可,设\(P=\begin{bmatrix} b_{11}& b_{12}& ...& b_{1s}\\ b_{21}& b_{22}& ...& b_{2s}\\ :& :& :& :\\ b_{s1}& b_{s2}& ...&b_{ss} \end{bmatrix}\),则
\((α_{1},α_{2},...,α_{s})P\)
\(=(α_{1},α_{2},...,α_{s})\begin{bmatrix} b_{11}& b_{12}& ...& b_{1s}\\ b_{21}& b_{22}& ...& b_{2s}\\ :& :& :& :\\ b_{s1}& b_{s2}& ...&b_{ss} \end{bmatrix}\),因此
\(\begin{equation} \begin{aligned} &f((α_{1},α_{2},...,α_{s})P)\\ =&f((b_{11}+b_{21}+...+b_{s1})α_{1}\\ &+(b_{12}+b_{22}+...+b_{s2})α_{2}\\ &+...+(b_{1s}+b_{2s}+...+b_{ss})α_{s}) \end{aligned} \end{equation}\)
由于\(f\)为线性变换,故
\(f((α_{1},α_{2},...,α_{s})P)\)
\(=(b_{11}+b_{21}+...+b_{s1})f(α_{1})\)
\(+(b_{12}+b_{22}+...+b_{s2})f(α_{2})\)
\(+...+(b_{1s}+b_{2s}+...+b_{ss})f(α_{s})\)
\(=(f(α_{1}),f(α_{2}),...,f(α_{s}))P\)
,对于\(f(α_{1},α_{2},...,α_{s})\)
\(=(f(α_{1}),f(α_{2}),...,f(α_{s}))\),则只是一个定义问题,
之所以这样定义,是因为比起\((f(α_{1}),f(α_{2}),...,f(α_{s}))\),
\(f(α_{1},α_{2},...,α_{s})\)的书写形式更加简洁,这样定义以后
\(f((α_{1},α_{2},...,α_{s})X)\)
\(=f(α_{1},α_{2},...,α_{s})X\),其中\(X\)为任意\(s\)阶方阵
实际上,若将\((α_{1},α_{2},...,α_{s})\)视为是一个数阵,将\(P\)中的列向量视为一组基,也即对\(P\)作线性变换,显然对\(P\)作线性变换
时,\(f\)在以\(P\)的列向量为基时的矩阵与以\(α_{1},α_{2},...,α_{s}\)为基时的矩阵不同,即同一线性变换在不同基下的矩阵一般是不同的

浙公网安备 33010602011771号