Multi-Cell Downlink Beamforming: Direct FP, Closed-Form FP, Weighted MMSE

这里写自定义目录标题

  • Direct FP
  • Closed-Form FP
    • the Lagrangian function
    • the Lagrange dual function: maximizing the Lagrangian
    • the Lagrange dual problem: minimizing the Lagrange dual function
    • Closed-Form FP
  • Weighted MMSE
    • 原论文
  • Lagrange dual
    • 5.1.1 The Lagrangian
    • 5.1.2 The Lagrange dual function
    • 5.2 The Lagrange dual problem
    • 5.2.3 Strong duality and Slater’s constraint qualification
    • 5.2.3 Strong duality and Slater’s constraint qualification
    • 5.5.3 KKT optimality conditions
  • 仿真

Multi-User in each Cell, MISO
沈闓明代码

Direct FP

K. Shen and W. Yu, “Fractional Programming for Communication Systems—Part I: Power Control and Beamforming,” in IEEE Transactions on Signal Processing, vol. 66, no. 10, pp. 2616-2630, 15 May15, 2018, doi: 10.1109/TSP.2018.2812733.

the multidimensional quadratic transform

∑ n = 1 N ∑ k = 1 K log ⁡ 2 ( 1 + ∣ h n , n , k H w n , k ∣ 2 ∑ ( j , i ) ≠ ( n , k ) ∣ h j , n , k H w j , i ∣ 2 + σ n , k 2 ) \sum\limits_{n = 1}^N {\sum\limits_{k = 1}^K {{{\log }_2}\left( {1 + \frac{{{{\left| {{\bf{h}}_{n,n,k}^H{{\bf{w}}_{n,k}}} \right|}^2}}}{{\sum\limits_{(j,i) \ne (n,k)} {{{\left| {{\bf{h}}_{j,n,k}^H{{\bf{w}}_{j,i}}} \right|}^2}} + \sigma _{n,k}^2}}} \right)} } n=1Nk=1Klog2 1+(j,i)=(n,k)hj,n,kHwj,i2+σn,k2hn,n,kHwn,k2
the direct FP approach applies the multidimensional quadratic transform (Theorem 2) to each SINR term.
f q ( W , Y ) = ∑ ( n , k ) log ⁡ ( 1 + 2 R e { y n , k H w n , k H h n , n , k } − ∣ y n , k ∣ 2 ( ∑ ( j , i ) ≠ ( n , k ) ∣ h j , n , k H w j , i ∣ 2 + σ n , k 2 ) ) {f_q}\left( {{\bf{W}},{\bf{Y}}} \right) = \sum\limits_{(n,k)} {\log \left( {1 + 2{\rm{Re}}\left\{ {y_{n,k}^H{\bf{w}}_{n,k}^H{{\bf{h}}_{n,n,k}}} \right\} - {{\left| {{y_{n,k}}} \right|}^2}\left( {\sum\limits_{(j,i) \ne (n,k)} {{{\left| {{\bf{h}}_{j,n,k}^H{{\bf{w}}_{j,i}}} \right|}^2}} + \sigma _{n,k}^2} \right)} \right)} fq(W,Y)=(n,k)log(1+2Re{yn,kHwn,kHhn,n,k}yn,k2((j,i)=(n,k) hj,n,kHwj,i 2+σn,k2))

Direct FP
初始化 w n , k , ∀ n , k {\bf{w}}_{n,k}, \forall n,k wn,k,n,k
重复

  1. 更新 y n , k ⋆ = h n , n , k H w n , k ∑ ( j , i ) ≠ ( n , k ) ∣ h j , n , k H w j , i ∣ 2 + σ n , k 2 y_{n,k}^ \star = \frac{{{\bf{h}}_{n,n,k}^H{{\bf{w}}_{n,k}}}}{{\sum\limits_{(j,i) \ne (n,k)} {{{\left| {{\bf{h}}_{j,n,k}^H{{\bf{w}}_{j,i}}} \right|}^2}} + \sigma _{n,k}^2}} yn,k=(j,i)=(n,k)hj,n,kHwj,i2+σn,k2hn,n,kHwn,k with w n , k , ∀ n , k {\bf{w}}_{n,k}, \forall n,k wn,k,n,k
  2. 给定 y n , k {y_{n,k}} yn,k,求解问题,更新 w n , k {{\bf{w}}_{n,k}} wn,k
    max ⁡ { w n , k , y n , k } f q ( W , Y ) s . t . ∑ k = 1 K w n , k H w n , k ≤ p ˉ n , ∀ n = 1 , … , N , \begin{array}{l} \mathop {\max }\limits_{\left\{ {{{\bf{w}}_{n,k}},{y_{n,k}}} \right\}} \;\;{f_q}\left( {{\bf{W}},{\bf{Y}}} \right)\\ {\rm{s}}.{\rm{t}}.\;\sum\limits_{k = 1}^K {{\bf{w}}_{n,k}^H{{\bf{w}}_{n,k}}} \le {{\bar p}_n},\forall n = 1, \ldots ,N, \end{array} {wn,k,yn,k}maxfq(W,Y)s.t.k=1Kwn,kHwn,kpˉn,n=1,,N,
    the optimization problem is a convex problem of w n , k {{\bf{w}}_{n,k}} wn,k when the auxiliary variable y n , k {y_{n,k}} yn,k is held fixed.

直到 f q ( W , Y ) {f_q}\left( {{\bf{W}},{\bf{Y}}} \right) fq(W,Y)收敛

Closed-Form FP

K. Shen and W. Yu, “Fractional Programming for Communication Systems—Part I: Power Control and Beamforming,” in IEEE Transactions on Signal Processing, vol. 66, no. 10, pp. 2616-2630, 15 May15, 2018, doi: 10.1109/TSP.2018.2812733.

∑ n = 1 N ∑ k = 1 K log ⁡ 2 ( 1 + ∣ h n , n , k H w n , k ∣ 2 ∑ ( j , i ) ≠ ( n , k ) ∣ h j , n , k H w j , i ∣ 2 + σ n , k 2 ) \sum\limits_{n = 1}^N {\sum\limits_{k = 1}^K {{{\log }_2}\left( {1 + \frac{{{{\left| {{\bf{h}}_{n,n,k}^H{{\bf{w}}_{n,k}}} \right|}^2}}}{{\sum\limits_{(j,i) \ne (n,k)} {{{\left| {{\bf{h}}_{j,n,k}^H{{\bf{w}}_{j,i}}} \right|}^2}} + \sigma _{n,k}^2}}} \right)} } n=1Nk=1Klog2 1+(j,i)=(n,k)hj,n,kHwj,i2+σn,k2hn,n,kHwn,k2
Lagrangian Dual Transform (Multidimensional and Complex)
f r ( W , U ) = ∑ ( n , k ) ( log ⁡ ( 1 + u n , k ) − u n , k + ( 1 + u n , k ) ∣ h n , n , k H w n , k ∣ 2 ∑ ( j , i ) ∣ h j , n , k H w j , i ∣ 2 + σ n , k 2 ) {f_r}\left( {{\bf{W}},{\bf{U}}} \right) = \sum\limits_{(n,k)} {\left( {\log \left( {1 + {u_{n,k}}} \right) - {u_{n,k}} + \left( {1 + {u_{n,k}}} \right)\frac{{{{\left| {{\bf{h}}_{n,n,k}^H{{\bf{w}}_{n,k}}} \right|}^2}}}{{\sum\limits_{(j,i)} {{{\left| {{\bf{h}}_{j,n,k}^H{{\bf{w}}_{j,i}}} \right|}^2}} + \sigma _{n,k}^2}}} \right)} fr(W,U)=(n,k) log(1+un,k)un,k+(1+un,k)(j,i)hj,n,kHwj,i2+σn,k2hn,n,kHwn,k2
∂ ∂ u n , k f r ( W , U ) = 0 \frac{\partial }{{\partial {u_{n,k}}}}{f_r}\left( {{\bf{W}},{\bf{U}}} \right) = 0 un,kfr(W,U)=0
u n , k ⋆ = γ n , k = ∣ h n , n , k H w n , k ∣ 2 ∑ ( j , i ) ≠ ( n , k ) ∣ h j , n , k H w j , i ∣ 2 + σ n , k 2 ∈ R 1 u_{n,k}^ \star = {\gamma _{n,k}} = \frac{{{{\left| {{\bf{h}}_{n,n,k}^H{{\bf{w}}_{n,k}}} \right|}^2}}}{{\sum\limits_{(j,i) \ne (n,k)} {{{\left| {{\bf{h}}_{j,n,k}^H{{\bf{w}}_{j,i}}} \right|}^2}} + \sigma _{n,k}^2}} \in {{\mathbb{R}}^1} un,k=γn,k=(j,i)=(n,k)hj,n,kHwj,i2+σn,k2hn,n,kHwn,k2R1
Quadratic Transform (Multidimensional)
f q ( W , U , V ) = ∑ ( n , k ) ( 2 ( 1 + u n , k ) R e { w n , k H h n , n , k v n , k } − ∣ v n , k ∣ 2 ( ∑ ( j , i ) ∣ h j , n , k H w j , i ∣ 2 + σ n , k 2 ) ) + c o n s t ( U ) {f_q}\left( {{\bf{W}},{\bf{U}},{\bf{V}}} \right) = \sum\limits_{(n,k)} {\left( {2\sqrt {(1 + {u_{n,k}})} {\rm{Re}}\left\{ {{\bf{w}}_{n,k}^H{{\bf{h}}_{n,n,k}}{v_{n,k}}} \right\} - {{\left| {{v_{n,k}}} \right|}^2}\left( {\sum\limits_{(j,i)} {{{\left| {{\bf{h}}_{j,n,k}^H{{\bf{w}}_{j,i}}} \right|}^2}} + \sigma _{n,k}^2} \right)} \right)} + {\rm{const}}({\bf{U}}) fq(W,U,V)=(n,k)(2(1+un,k) Re{wn,kHhn,n,kvn,k}vn,k2((j,i) hj,n,kHwj,i 2+σn,k2))+const(U)
∂ ∂ v n , k f q ( W , U , V ) = 0 \frac{\partial }{{\partial {v_{n,k}}}}{f_q}\left( {{\bf{W}},{\bf{U}},{\bf{V}}} \right) = 0 vn,kfq(W,U,V)=0
v n , k ⋆ = ( 1 + u n , k ) h n , n , k H w n , k ∑ ( j , i ) ∣ h j , n , k H w j , i ∣ 2 + σ n , k 2 v_{n,k}^ \star = \frac{{\sqrt {(1 + {u_{n,k}})} {\bf{h}}_{n,n,k}^H{{\bf{w}}_{n,k}}}}{{\sum\limits_{(j,i)} {{{\left| {{\bf{h}}_{j,n,k}^H{{\bf{w}}_{j,i}}} \right|}^2}} + \sigma _{n,k}^2}} vn,k=(j,i)hj,n,kHwj,i2+σn,k2(1+un,k) hn,n,kHwn,k

Transformed Problem
max ⁡ { w n , k } f q ( W , U , V ) s . t . p ˉ n − ∑ k = 1 K w n , k H w n , k ≥ 0 , ∀ n = 1 , … , N , \begin{array}{l} \mathop {\max }\limits_{\left\{ {{{\bf{w}}_{n,k}}} \right\}} \;\;{f_q}\left( {{\bf{W}},{\bf{U}},{\bf{V}}} \right)\\ {\rm{s}}.{\rm{t}}.\;{{\bar p}_n} - \sum\limits_{k = 1}^K {{\bf{w}}_{n,k}^H{{\bf{w}}_{n,k}}} \ge 0,\forall n = 1, \ldots ,N, \end{array} {wn,k}maxfq(W,U,V)s.t.pˉnk=1Kwn,kHwn,k0,n=1,,N,

the Lagrangian function

L ( W , U , V , η ) = f q ( W , U , V ) + ∑ n = 1 N η n ( p ˉ n − ∑ k = 1 K w n , k H w n , k ) L({\bf{W}},{\bf{U}},{\bf{V}},{\bm{\eta}}) = {f_q}({\bf{W}},{\bf{U}},{\bf{V}}) + \sum\limits_{n = 1}^N {{\eta _n}\left( {{{\bar p}_n} - \sum\limits_{k = 1}^K {{\bf{w}}_{n,k}^H{{\bf{w}}_{n,k}}} } \right)} L(W,U,V,η)=fq(W,U,V)+n=1Nηn(pˉnk=1Kwn,kHwn,k)

the Lagrange dual function: maximizing the Lagrangian

g ( η ) = m a x { w n , k } L ( W , U , V , η ) g\left( {\bm{\eta}} \right) = \mathop {{\rm{max}}}\limits_{\left\{ {{{\bf{w}}_{n,k}}} \right\}} \;L({\bf{W}},{\bf{U}},{\bf{V}},{\bm{\eta}}) g(η)={wn,k}maxL(W,U,V,η)
∂ ∂ w n , k L ( W , U , V , η ) = 0 ⇒ \frac{\partial }{{\partial {{\bf{w}}_{n,k}}}}L({\bf{W}},{\bf{U}},{\bf{V}},{\bm{\eta}}) = 0 \Rightarrow wn,kL(W,U,V,η)=0
w n , k ∗ = ( 1 + u n , k ) h n , n , k v n , k ∑ ( m , l ) ( h n , m , l v m , l v m , l H h n , m , l H ) + η n I {\bf{w}}_{n,k}^* = \frac{{\sqrt {(1 + {u_{n,k}})} {{\bf{h}}_{n,n,k}}{v_{n,k}}}}{{\sum\limits_{(m,l)} {\left( {{{\bf{h}}_{n,m,l}}{v_{m,l}}v_{m,l}^H{\bf{h}}_{n,m,l}^H} \right)} + {\eta _n}I}} wn,k=(m,l)(hn,m,lvm,lvm,lHhn,m,lH)+ηnI(1+un,k) hn,n,kvn,k

the Lagrange dual problem: minimizing the Lagrange dual function

Lagrange multipliers are component-wise non-negative
m i n η ≥ 0 g ( η ) \mathop {{\rm{min}}}\limits_{{\bm{\eta}} \ge 0} g\left( {\bm{\eta}} \right) η0ming(η)

Closed-Form FP

Closed-Form FP
初始化 w n , k , ∀ n , k {\bf{w}}_{n,k}, \forall n,k wn,k,n,k
重复

  1. 更新 u n , k ⋆ = γ n , k = ∣ h n , n , k H w n , k ∣ 2 ∑ ( j , i ) ≠ ( n , k ) ∣ h j , n , k H w j , i ∣ 2 + σ n , k 2 ∈ R 1 u_{n,k}^ \star = {\gamma _{n,k}} = \frac{{{{\left| {{\bf{h}}_{n,n,k}^H{{\bf{w}}_{n,k}}} \right|}^2}}}{{\sum\limits_{(j,i) \ne (n,k)} {{{\left| {{\bf{h}}_{j,n,k}^H{{\bf{w}}_{j,i}}} \right|}^2}} + \sigma _{n,k}^2}} \in {{\mathbb{R}}^1} un,k=γn,k=(j,i)=(n,k)hj,n,kHwj,i2+σn,k2hn,n,kHwn,k2R1 with w n , k , ∀ n , k {\bf{w}}_{n,k}, \forall n,k wn,k,n,k
  2. 更新 v n , k ⋆ = ( 1 + u n , k ) h n , n , k H w n , k ∑ ( j , i ) ∣ h j , n , k H w j , i ∣ 2 + σ n , k 2 v_{n,k}^ \star = \frac{{\sqrt {(1 + {u_{n,k}})} {\bf{h}}_{n,n,k}^H{{\bf{w}}_{n,k}}}}{{\sum\limits_{(j,i)} {{{\left| {{\bf{h}}_{j,n,k}^H{{\bf{w}}_{j,i}}} \right|}^2}} + \sigma _{n,k}^2}} vn,k=(j,i)hj,n,kHwj,i2+σn,k2(1+un,k) hn,n,kHwn,k with w n , k , ∀ n , k {\bf{w}}_{n,k}, \forall n,k wn,k,n,k and u n , k , ∀ n , k u_{n,k}, \forall n,k un,k,n,k
  3. 更新 η n , ∀ n {\eta _n}, \forall n ηn,n。利用二分法可以找到最小的 η n {\eta _n} ηn ,使得 ∑ k = 1 K w n , k H ( η n ) w n , k ( η n ) = p ˉ n \sum\limits_{k = 1}^K {{\bf{w}}_{n,k}^H\left( {{\eta _n}} \right){{\bf{w}}_{n,k}}\left( {{\eta _n}} \right)} = {\bar p_n} k=1Kwn,kH(ηn)wn,k(ηn)=pˉn
    其中, w n , k ( η n ) = ( 1 + u n , k ) h n , n , k v n , k ∑ ( m , l ) ( h n , m , l v m , l v m , l H h n , m , l H ) + η n I {{\bf{w}}_{n,k}}\left( {{\eta _n}} \right) = \frac{{\sqrt {(1 + {u_{n,k}})} {{\bf{h}}_{n,n,k}}{v_{n,k}}}}{{\sum\limits_{(m,l)} {\left( {{{\bf{h}}_{n,m,l}}{v_{m,l}}v_{m,l}^H{\bf{h}}_{n,m,l}^H} \right)} + {\eta _n}I}} wn,k(ηn)=(m,l)(hn,m,lvm,lvm,lHhn,m,lH)+ηnI(1+un,k) hn,n,kvn,k
  4. 更新 w n , k ∗ = ( 1 + u n , k ) h n , n , k v n , k ∑ ( m , l ) ( h n , m , l v m , l v m , l H h n , m , l H ) + η n I {\bf{w}}_{n,k}^* = \frac{{\sqrt {(1 + {u_{n,k}})} {{\bf{h}}_{n,n,k}}{v_{n,k}}}}{{\sum\limits_{(m,l)} {\left( {{{\bf{h}}_{n,m,l}}{v_{m,l}}v_{m,l}^H{\bf{h}}_{n,m,l}^H} \right)} + {\eta _n}I}} wn,k=(m,l)(hn,m,lvm,lvm,lHhn,m,lH)+ηnI(1+un,k) hn,n,kvn,k with u n , k , ∀ n , k u_{n,k}, \forall n,k un,k,n,k, v n , k , ∀ n , k v_{n,k}, \forall n,k vn,k,n,k, and η n , ∀ n {\eta _n}, \forall n ηn,n

直到 f q ( W , U , V ) {f_q}\left( {{\bf{W}},{\bf{U}},{\bf{V}}} \right) fq(W,U,V)收敛

对偶变量或Lagrange multipliers的更新:KKT条件 η n ( p ˉ n − ∑ k = 1 K w n , k H w n , k ) = 0 , ∀ n = 1 , … , N , {\eta _n}\left( {{{\bar p}_n} - \sum\limits_{k = 1}^K {{\bf{w}}_{n,k}^H{{\bf{w}}_{n,k}}} } \right) = 0,\forall n = 1, \ldots ,N, ηn(pˉnk=1Kwn,kHwn,k)=0,n=1,,N,
∑ k = 1 K w n , k H ( η n ) w n , k ( η n ) \sum\limits_{k = 1}^K {{\bf{w}}_{n,k}^H\left( {{\eta _n}} \right){{\bf{w}}_{n,k}}\left( {{\eta _n}} \right)} k=1Kwn,kH(ηn)wn,k(ηn)是关于 η n {\eta _n} ηn的单调递减函数,利用二分法可以找到最小的 η n {\eta _n} ηn ,使得 ∑ k = 1 K w n , k H ( η n ) w n , k ( η n ) = p ˉ n \sum\limits_{k = 1}^K {{\bf{w}}_{n,k}^H\left( {{\eta _n}} \right){{\bf{w}}_{n,k}}\left( {{\eta _n}} \right)} = {\bar p_n} k=1Kwn,kH(ηn)wn,k(ηn)=pˉn
η n < η n ∗ {\eta _n} < \eta _n^* ηn<ηn时, ∑ k = 1 K w n , k H ( η n ) w n , k ( η n ) \sum\limits_{k = 1}^K {{\bf{w}}_{n,k}^H\left( {{\eta _n}} \right){{\bf{w}}_{n,k}}\left( {{\eta _n}} \right)} k=1Kwn,kH(ηn)wn,k(ηn) >基站最大发射功率 p ˉ n {\bar p_n} pˉn
η n = η n ∗ {\eta _n} = \eta _n^* ηn=ηn 时, ∑ k = 1 K w n , k H ( η n ) w n , k ( η n ) \sum\limits_{k = 1}^K {{\bf{w}}_{n,k}^H\left( {{\eta _n}} \right){{\bf{w}}_{n,k}}\left( {{\eta _n}} \right)} k=1Kwn,kH(ηn)wn,k(ηn) =基站最大发射功率 p ˉ n {\bar p_n} pˉn
η n > η n ∗ {\eta _n} > \eta _n^* ηn>ηn时, ∑ k = 1 K w n , k H ( η n ) w n , k ( η n ) \sum\limits_{k = 1}^K {{\bf{w}}_{n,k}^H\left( {{\eta _n}} \right){{\bf{w}}_{n,k}}\left( {{\eta _n}} \right)} k=1Kwn,kH(ηn)wn,k(ηn) <基站最大发射功率 p ˉ n {\bar p_n} pˉn

Weighted MMSE

u n , k = h n , n , k H w n , k ∑ ( m , j ) ∣ h m , n , k H w m , j ∣ 2 + σ n , k 2 {u_{n,k}} = \frac{{{\bf{h}}_{n,n,k}^H{{\bf{w}}_{n,k}}}}{{\sum\limits_{(m,j)} {{{\left| {{\bf{h}}_{m,n,k}^H{{\bf{w}}_{m,j}}} \right|}^2}} + \sigma _{n,k}^2}} un,k=(m,j)hm,n,kHwm,j2+σn,k2hn,n,kHwn,k
(图中h的下标打错了)

hm,n,k denote the downlink channel between BS m and UE k in cell n
wn,k denote the beamformer for UE k in cell n
the optimum Lagrange multiplier μ n ⋆ \mu _n^ \star μn can be determined efficiently by a bisection search method.
Weighted MMSE

原论文

Q. Shi, M. Razaviyayn, Z. -Q. Luo and C. He, “An Iteratively Weighted MMSE Approach to Distributed Sum-Utility Maximization for a MIMO Interfering Broadcast Channel,” in IEEE Transactions on Signal Processing, vol. 59, no. 9, pp. 4331-4340, Sept. 2011, doi: 10.1109/TSP.2011.2147784.

Weighted MMSE
V i k {{\bf{V}}_{{i_k}}} Vik 表示基站k对用户 i k {i_k} ik 的波束成形
H i k , j {{\bf{H}}_{{i_k},j}} Hik,j 表示从基站j到用户 i k {i_k} ik的信道
u k , i = h k , k , i H w k , i ∑ ( j , l ) ∣ h j , k , i H w j , l ∣ 2 + σ k , i 2 {u_{k,i}} = \frac{{{\bf{h}}_{k,k,i}^H{{\bf{w}}_{k,i}}}}{{\sum\limits_{(j,l)} {{{\left| {{\bf{h}}_{j,k,i}^H{{\bf{w}}_{j,l}}} \right|}^2}} + \sigma _{k,i}^2}} uk,i=(j,l)hj,k,iHwj,l2+σk,i2hk,k,iHwk,i

Lagrange dual

上海交通大学 CS257 Linear and Convex Optimization
南京大学 Duality (I) - NJU

the standard form (5.1)
在这里插入图片描述
min ⁡ X f ( X ) s . t . g i ( X ) ≤ 0 , ∀ i = 1 , … , m , \begin{array}{l} {\mathop {\min }_{\bf{X}} \;\;f\left( {\bf{X}} \right)}\\ {{\rm{s}}.{\rm{t}}.\;{g_i}\left( {\bf{X}} \right) \le 0,\forall i = 1, \ldots ,m,} \end{array} minXf(X)s.t.gi(X)0,i=1,,m,

5.1.1 The Lagrangian

在这里插入图片描述
the dual variables or Lagrange multiplier vectors associated with the problem (5.1).

5.1.2 The Lagrange dual function

the minimum value of the Lagrangian
在这里插入图片描述

5.2 The Lagrange dual problem

the Lagrange dual problem associated with the problem (5.1).
在这里插入图片描述

5.2.3 Strong duality and Slater’s constraint qualification

在这里插入图片描述

5.2.3 Strong duality and Slater’s constraint qualification

Slater’s theorem states that
strong duality holds, if Slater’s condition holds (and the problem is convex).
strong duality obtains, when the primal problem is convex and Slater’s condition holds

Slater’s Condition for Convex Problems
上海交通大学 CS257 Linear and Convex Optimization
在这里插入图片描述

5.5.3 KKT optimality conditions

Karush-Kuhn-Tucker (KKT) conditions
在这里插入图片描述
for any optimization problem with differentiable objective and constraint functions for which strong duality obtains, any pair of primal and dual optimal points must satisfy the KKT conditions (5.49).

在这里插入图片描述

仿真

本文来自互联网用户投稿,该文观点仅代表作者本人,不代表本站立场。本站仅提供信息存储空间服务,不拥有所有权,不承担相关法律责任。如若转载,请注明出处:http://www.mzph.cn/news/206870.shtml

如若内容造成侵权/违法违规/事实不符,请联系多彩编程网进行投诉反馈email:809451989@qq.com,一经查实,立即删除!

相关文章

阿里云服务器经济型、通用算力型、计算型、通用型、内存型实例区别及选择参考

当我们通过阿里云的活动购买云服务器会发现&#xff0c;相同配置的云服务器往往有多个不同的实例可选&#xff0c;而且价格差别也比较大&#xff0c;例如同样是4核8G的配置的云服务器&#xff0c;经济型e实例活动价格只要1500.48/1年起&#xff0c;通用算力型u1实例要1795.97/1…

nvidia安装出现7-zip crc error解决办法

解决办法&#xff1a;下载network版本&#xff0c;重新安装。&#xff08;选择自己需要的版本&#xff09; 网址&#xff1a;CUDA Toolkit 12.3 Update 1 Downloads | NVIDIA Developer 分析原因&#xff1a;local版本的安装包可能在下载过程中出现损坏。 本人尝试过全网说的…

linux 系统安全基线 安全加固操作

目录 用户口令设置 root用户远程登录限制 检查是否存在除root之外UID为0的用户 ​​​​​​​root用户环境变量的安全性 ​​​​​​​远程连接的安全性配置 ​​​​​​​用户的umask安全配置 ​​​​​​​重要目录和文件的权限设置 ​​​​​​​找未授权的SUID…

json转yolo格式

json转yolo格式 视觉分割得一些标注文件是json格式&#xff0c;比如&#xff0c;舌头将这个舌头区域分割出来&#xff08;用mask二值图的形式&#xff09;&#xff0c;对舌头的分割第一步是需要检测出来&#xff0c;缺少数据集&#xff0c;可以使用分割出来的结果&#xff0c;将…

无公网IP环境如何SSH远程连接Deepin操作系统

文章目录 前言1. 开启SSH服务2. Deppin安装Cpolar3. 配置ssh公网地址4. 公网远程SSH连接5. 固定连接SSH公网地址6. SSH固定地址连接测试 前言 Deepin操作系统是一个基于Debian的Linux操作系统&#xff0c;专注于使用者对日常办公、学习、生活和娱乐的操作体验的极致&#xff0…

数据仪表盘设计:可视化数据指标和报告

写在开头 在信息爆炸的时代,数据不再是简单的数字和图表,而是一种有机的信息体系。如何将这些琳琅满目的数据以一种直观而高效的方式展示,成为企业决策者和分析师们共同关注的问题。本文将带您深入学习如何设计和创建数据仪表盘,使数据指标和报告以一目了然的方式呈现。 …

Python---time库

目录 时间获取 时间格式化 程序计时 time库包含三类函数&#xff1a; 时间获取&#xff1a;time() ctime() gmtime() 时间格式化&#xff1a;strtime() strptime() 程序计时&#xff1a;sleep() perf_counter() 下面逐一介绍&#…

H3.3K27M弥漫性中线胶质瘤的反义寡核苷酸治疗

今天给同学们分享一篇实验文章“Antisense oligonucleotide therapy for H3.3K27M diffuse midline glioma”&#xff0c;这篇文章发表在Sci Transl Med期刊上&#xff0c;影响因子为17.1。 结果解读&#xff1a; CRISPR-Cas9消耗H3.3K27M恢复了H3K27三甲基化&#xff0c;并延…

Echarts地图案例及常见问题

前言 ECharts 是一个使用 JavaScript 实现的开源可视化库,它可以帮助用户以简单的方式创建复杂的时间序列、条形图、饼图、地图等图形。 Echarts绘制地图的案例 展示了中国各省份的人口数量 var myChart = echarts.init(document.getElementById(main)); var option = {t…

【TailwindCSS】

TailwindCSS作为一种现代化的CSS框架&#xff0c;以其高度的定制性和灵活性受到前端开发者的青睐。本文旨在提供一份详细的TailwindCSS使用教程&#xff0c;特别适用于Vite和Vue框架的组合。 我们将从安装开始&#xff0c;深入探讨如何在项目中有效利用TailwindCSS的各项功能&…

在AWS Lambda上部署标准FFmpeg工具——Docker方案

大纲 1 确定Lambda运行时环境1.1 Lambda系统、镜像、内核版本1.2 运行时1.2.1 Python1.2.2 Java 2 启动EC23 编写调用FFmpeg的代码4 生成docker镜像4.1 安装和启动Docker服务4.2 编写Dockerfile脚本4.3 生成镜像 5 推送镜像5.1 创建存储库5.2 给EC2赋予角色5.2.1 创建策略5.2.2…

【带头学C++】----- 九、类和对象 ---- 9.10 C++设计模式之单例模式设计

❤️❤️❤️❤️❤️❤️❤️❤️❤️❤️❤️麻烦您点个关注&#xff0c;不迷路❤️❤️❤️❤️❤️❤️❤️❤️❤️❤️❤️❤️ 目 录 9.10 C设计模式之单例模式设计 举例说明&#xff1a; 9.10 C设计模式之单例模式设计 看过我之前的文章的&#xff0c;简单讲解过C/Q…

遥测终端机RTU:实现远程监测和控制的重要工具

遥测终端机RTU对设备进行远程监测和控制&#xff0c;支持采集和传输数据&#xff0c;以实现对工业过程、公用事业、水文和环境的监测和管理。 遥测终端机RTU工作原理 计讯物联遥测终端机RTU通过网口、串口进行传感器/设备等现场数据采集&#xff0c;将其转换为数字信号&#xf…

【LeetCode】202. 快乐数

202. 快乐数 难度&#xff1a;简单 题目 编写一个算法来判断一个数 n 是不是快乐数。 「快乐数」 定义为&#xff1a; 对于一个正整数&#xff0c;每一次将该数替换为它每个位置上的数字的平方和。然后重复这个过程直到这个数变为 1&#xff0c;也可能是 无限循环 但始终变…

高校网站建设的效果如何

高校有较高的信息承载需求、招生宣传、学校内容呈现、内部消息触达等需求&#xff0c;对高校来说&#xff0c;如今互联网深入生活各个场景&#xff0c;无论学校发展、外部拓展还是内部师生互动、通知触达等都需要完善。 除了传统传单及第三方平台展示外&#xff0c;学校构建属…

C#-数组池减少GC工作

数组池减少GC工作 通过ArrayPool类&#xff08;名称空间System.Buffers&#xff09;使用数组池&#xff0c;可减少垃圾收集器的工作&#xff0c;ArrayPool管理一个数组池&#xff0c;数组可以从这租借&#xff0c;并返回池中&#xff0c;内存在ArrayPool中管理。 创建ArrayPool…

Html5响应式全开源网站建站源码系统 附带完整的搭建教程

Html5响应式全开源网站建站源码系统是基于Html5、CSS3和JavaScript等技术开发的全开源网站建站系统。它旨在为初学者和小型企业提供一套快速、简便的网站建设解决方案。该系统采用响应式设计&#xff0c;可以自适应不同设备的屏幕大小&#xff0c;提高用户体验。同时&#xff0…

Clean My Mac X2024解锁完整版本

Clean My Mac X是Mac上一款美观易用的系统优化清理工具&#xff0c;也是小编刚开始用Mac时的装机必备。垃圾需要时时清&#xff0c;电脑才能常年新。Windows的垃圾清理工具选择有很多&#xff0c;但是Mac的清理工具可选择的就很少。 今天给大家推荐大名鼎鼎的Clean My Mac X&a…

elasticsearch-head 启动教程

D:\elasticsearch-head-master>grunt server ‘grunt’ 不是内部或外部命令&#xff0c;也不是可运行的程序 或批处理文件。 npm install -g grunt-clinpm install

Leetcode—190.颠倒二进制位【简单】

2023每日刷题&#xff08;五十二&#xff09; Leetcode—190.颠倒二进制位 算法思路 实现代码 class Solution { public:uint32_t reverseBits(uint32_t n) {uint32_t res 0;for(int i 0; i < 32 && n > 0; i) {res | (n & 1) << (31 - i);n >&…