矩阵的迹(trace)

定义

n × n \mathit{n}\times \mathit{n} n×n矩阵 A \boldsymbol A A的对角元素之和称为 A \boldsymbol A A的迹( t r a c e \mathit{trace} trace),记作 t r ( A ) \mathit{tr}(\boldsymbol A) tr(A),即有 t r ( A ) = a 11 + ⋯ + a n n = ∑ i = 1 n a i i \mathit{tr}(\boldsymbol A)=\mathit{a}_{11} +\cdots +\mathit{a}_{\mathit{nn}}=\sum_{\mathit{i}=1}^{\mathit{n}} \mathit{a}_{\mathit{ii}} tr(A)=a11++ann=i=1naii

迹的等式性质

  1. A \boldsymbol A A B \boldsymbol B B均为 n × n \mathit{n}\times \mathit{n} n×n矩阵,则 t r ( A ± B ) = t r ( A ) ± t r ( B ) \mathit{tr}(\boldsymbol A\pm \boldsymbol B)=\mathit{tr}(\boldsymbol A)\pm\mathit{tr}(\boldsymbol B) tr(A±B)=tr(A)±tr(B)
  2. A \boldsymbol A A B \boldsymbol B B均为 n × n \mathit{n}\times \mathit{n} n×n矩阵,且 c 1 \mathit{c}_{1} c1 c 2 \mathit{c}_{2} c2为常数,则 t r ( c 1 A ± c 2 B ) = c 1 t r ( A ) ± c 2 t r ( B ) \mathit{tr}(\mathit{c}_{1}\boldsymbol A\pm \mathit{c}_{2}\boldsymbol B)=\mathit{c}_{1}\mathit{tr}(\boldsymbol A)\pm\mathit{c}_{2}\mathit{tr}(\boldsymbol B) tr(c1A±c2B)=c1tr(A)±c2tr(B)
  3. 矩阵 A \boldsymbol A A的转置、共轭和复共轭转置的迹分别为 t r ( A T ) = t r ( A ) \mathit{tr} \left ( \boldsymbol {A}^{\mathrm{T} } \right ) =\mathit{tr}(\boldsymbol A) tr(AT)=tr(A) t r ( A ∗ ) = [ t r ( A ) ] ∗ \mathit{tr} \left ( \boldsymbol {A}^{\mathrm{*} } \right ) =[\mathit{tr} (\boldsymbol A)]^{*} tr(A)=[tr(A)] t r ( A H ) = [ t r ( A ) ] ∗ \mathit{tr} \left ( \boldsymbol {A}^{\mathrm{H} } \right ) =[\mathit{tr} (\boldsymbol A)]^{*} tr(AH)=[tr(A)]
  4. A ∈ C m × n \boldsymbol A\in {\Bbb C}^{\mathit{m} \times \mathit{n} } ACm×n B ∈ C n × m \boldsymbol B\in {\Bbb C}^{\mathit{n} \times \mathit{m} } BCn×m,则 t r ( A B ) = t r ( B A ) \mathit{tr} (\boldsymbol A\boldsymbol B)=\mathit{tr} (\boldsymbol B\boldsymbol A) tr(AB)=tr(BA)
  5. x H A x = t r ( A x x H ) \boldsymbol{x}^{\mathrm{H}} \boldsymbol{A} \boldsymbol{x}=\mathit{tr} \left(\boldsymbol{A} \boldsymbol{x} \boldsymbol{x}^{\mathrm{H}}\right) xHAx=tr(AxxH) y H x = t r ( x y H ) \boldsymbol{y}^{\mathrm{H}} \boldsymbol{x}=\mathit{tr} \left(\boldsymbol{x} \boldsymbol{y}^{\mathrm{H}}\right) yHx=tr(xyH)
  6. 迹等于特征值之和, t r ( A ) = λ 1 + ⋯ + λ n \mathit{tr} (\boldsymbol{A})=\mathit{\lambda} _{1}+\cdots+\mathit{\lambda} _{\mathit{n} } tr(A)=λ1++λn
  7. A ∈ C m × m \boldsymbol A\in {\Bbb C}^{\mathit{m} \times \mathit{m} } ACm×m B ∈ C m × n \boldsymbol B\in {\Bbb C}^{\mathit{m} \times \mathit{n} } BCm×n C ∈ C n × m \boldsymbol C\in {\Bbb C}^{\mathit{n} \times \mathit{m} } CCn×m D ∈ C n × n \boldsymbol D\in {\Bbb C}^{\mathit{n} \times \mathit{n} } DCn×n,分块矩阵的迹满足 t r [ A B C D ] = t r ( A ) + t r ( D ) \mathit{tr} \left[\begin{array}{ll} \boldsymbol{A} & \boldsymbol{B} \\ \boldsymbol{C} & \boldsymbol{D} \end{array}\right]=\mathit{tr}(\boldsymbol{A})+\mathit{tr}(\boldsymbol{D}) tr[ACBD]=tr(A)+tr(D)

应用迹的等式 t r ( A B ) = t r ( B A ) \mathit{tr} (\boldsymbol A\boldsymbol B)=\mathit{tr} (\boldsymbol B\boldsymbol A) tr(AB)=tr(BA),可以得到一些常用结论。例如,矩阵 A H A \boldsymbol{A}^{\mathrm{H} } \boldsymbol{A} AHA A A H \boldsymbol{A} \boldsymbol{A}^{\mathrm{H}} AAH的迹相等,有 t r ( A H A ) = t r ( A A H ) = ∑ i = 1 n ∑ j = 1 n a i j a i j ∗ = ∑ i = 1 n ∑ j = 1 n ∣ a i j ∣ 2 \mathit{tr} \left(\boldsymbol{A}^{\mathrm{H}} \boldsymbol{A}\right)=\mathit{tr}\left(\boldsymbol{A} \boldsymbol{A}^{\mathrm{H}}\right)=\sum_{\mathit{i}=1}^{\mathit{n}} \sum_{\mathit{j}=1}^{\mathit{n}} \mathit{a}_{\mathit{ij}} \mathit{a}_{\mathit{ij}}^{*}=\sum_{\mathit{i}=1}^{\mathit{n}} \sum_{\mathit{j}=1}^{\mathit{n}}\left|\mathit{a}_{\mathit{ij}}\right|^{2} tr(AHA)=tr(AAH)=i=1nj=1naijaij=i=1nj=1naij2类似地推广到多矩阵乘积的情形,有 t r ( A B C ) = t r ( B C A ) = t r ( C A B ) \mathit{tr} (\boldsymbol{A B C})=\mathit{tr} (\boldsymbol{B C A})=\mathit{tr} (\boldsymbol{C A B}) tr(ABC)=tr(BCA)=tr(CAB)以及 t r ( A B C D ) = t r ( B C D A ) = t r ( C D A B ) = t r ( D A B C ) \mathit{tr} (\boldsymbol{A B C D})=\mathit{tr}(\boldsymbol{B C D} \boldsymbol{A})=\mathit{tr}(\boldsymbol{C D} \boldsymbol{A} \boldsymbol{B})=\mathit{tr}(\boldsymbol{D} \boldsymbol{A B C}) tr(ABCD)=tr(BCDA)=tr(CDAB)=tr(DABC)

  • 2
    点赞
  • 2
    收藏
    觉得还不错? 一键收藏
  • 打赏
    打赏
  • 0
    评论

“相关推荐”对你有帮助么?

  • 非常没帮助
  • 没帮助
  • 一般
  • 有帮助
  • 非常有帮助
提交
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

打赏作者

Girl_We_Got_A

你的鼓励将是我创作的最大动力

¥1 ¥2 ¥4 ¥6 ¥10 ¥20
扫码支付:¥1
获取中
扫码支付

您的余额不足,请更换扫码支付或充值

打赏作者

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