MATLAB 数据索引问题。这里发生了什么?
MATLAB data indexing issue. What is going on here?
设 I
为恒等式,D
为正交投影,p
为向量。
我意识到我的几行代码合并为 (I-(I-D))(p)
,我可以将其简化为 D(p)
。在替换它时,我计算了新方法和旧方法,以仔细检查我是否在计算相同的东西(在我的代码的早期,我有一行是 D = I - D
。你在这里看到的 D
是D
.) 我没有得到相同的答案,并将其追溯到索引 D
.
中的错误
在这里你可以看到我正在使用调试器并检查 D 的部分并返回错误的数据。
右侧数据资源管理器中的值是我期望的值。有时我从 D(:,:,k,1)
得到了我期望的结果,有时我没有,即使我一个接一个地进行查询也是如此。
那些红色箭头指向的向量应该是相同的。在这些行之间没有任何其他更改或计算,并且当第一行是 运行 时 k = 2
。我关闭了 MATLAB 并重新启动它,每次都遇到同样的问题。 (D 取决于随机输入,但我没有改变种子,所以我在新打开 MATLAB 后第一次 运行 得到相同的东西。D
的计算方式,我 do 期望 D(:,:,1,1)
是单位矩阵。)
这世界到底是怎么回事?感谢任何帮助。
我想知道 MATLAB 是不是故意在搞我。有时当我打开它时,会弹出一个对话框说我需要更新我的学生许可证。我点击了更新按钮,但什么也没有发生,对话框也没有关闭,所以我点击了取消。
编辑:
K>> whos D P
Name Size Bytes Class Attributes
D 4-D 4608 double
P 4x1x6 192 double
K>> size(D)
ans =
4 4 6 6
我一直在研究 A
和 B
,我得到了同样的结果。有时它会正确计算,有时却不会。
K>> B=permute(P,[1,3,2])
B =
0.4155 0.27554 0.52338 0.6991 -0.11346 0.20999
0.53573 -0.83781 0.53182 -0.022364 0.60291 -0.62601
-0.49246 -0.46111 -0.39168 0.45919 0.42377 0.47074
0.54574 0.097595 0.53835 -0.54763 0.66637 0.58516
K>> A=D
A(:,:,1,1) =
1 0 0 0
0 1 0 0
0 0 1 0
0 0 0 1
A(:,:,2,1) =
0.99071 -0.091198 0.0020814 -0.029755
-0.091198 0.10503 0.020426 -0.292
0.0020814 0.020426 0.99953 0.0066643
-0.029755 -0.292 0.0066643 0.90473
A(:,:,3,1) =
0.46769 0.019281 -0.49725 0.036486
0.019281 0.9993 0.018011 -0.0013215
-0.49725 0.018011 0.53551 0.034083
0.036486 -0.0013215 0.034083 0.9975
A(:,:,4,1) =
0.96774 0.063488 -0.10826 0.12438
0.063488 0.87506 0.21304 -0.24477
-0.10826 0.21304 0.63673 0.41737
0.12438 -0.24477 0.41737 0.52047
A(:,:,5,1) =
0.7542 0.031217 0.42575 0.056052
0.031217 0.99604 -0.054071 -0.0071187
0.42575 -0.054071 0.26255 -0.097088
0.056052 -0.0071187 -0.097088 0.98722
A(:,:,6,1) =
0.9818 -0.10286 0.085279 0.0034902
-0.10286 0.41855 0.48208 0.01973
0.085279 0.48208 0.60031 -0.016358
0.0034902 0.01973 -0.016358 0.99933
A(:,:,1,2) =
0.99071 -0.091198 0.0020814 -0.029755
-0.091198 0.10503 0.020426 -0.292
0.0020814 0.020426 0.99953 0.0066643
-0.029755 -0.292 0.0066643 0.90473
A(:,:,2,2) =
1 0 0 0
0 1 0 0
0 0 1 0
0 0 0 1
A(:,:,3,2) =
0.97125 -0.15889 -0.0080537 -0.051131
-0.15889 0.12194 -0.044507 -0.28256
-0.0080537 -0.044507 0.99774 -0.014323
-0.051131 -0.28256 -0.014323 0.90907
A(:,:,4,2) =
0.91488 -0.16388 -0.18495 0.12967
-0.16388 0.6845 -0.35607 0.24964
-0.18495 -0.35607 0.59815 0.28174
0.12967 0.24964 0.28174 0.80247
A(:,:,5,2) =
0.95461 0.16812 0.10326 0.066372
0.16812 0.37733 -0.38244 -0.24582
0.10326 -0.38244 0.76511 -0.15098
0.066372 -0.24582 -0.15098 0.90295
A(:,:,6,2) =
0.99628 0.012018 0.052874 0.027665
0.012018 0.96117 -0.17085 -0.089393
0.052874 -0.17085 0.24833 -0.39329
0.027665 -0.089393 -0.39329 0.79422
A(:,:,1,3) =
0.46769 0.019281 -0.49725 0.036486
0.019281 0.9993 0.018011 -0.0013215
-0.49725 0.018011 0.53551 0.034083
0.036486 -0.0013215 0.034083 0.9975
A(:,:,2,3) =
0.97125 -0.15889 -0.0080537 -0.051131
-0.15889 0.12194 -0.044507 -0.28256
-0.0080537 -0.044507 0.99774 -0.014323
-0.051131 -0.28256 -0.014323 0.90907
A(:,:,3,3) =
1 0 0 0
0 1 0 0
0 0 1 0
0 0 0 1
A(:,:,4,3) =
0.98622 0.043449 -0.066709 0.085142
0.043449 0.86297 0.21038 -0.26852
-0.066709 0.21038 0.67698 0.41227
0.085142 -0.26852 0.41227 0.47382
A(:,:,5,3) =
0.62859 0.041458 0.47558 0.074661
0.041458 0.99537 -0.053085 -0.0083339
0.47558 -0.053085 0.39105 -0.0956
0.074661 -0.0083339 -0.0956 0.98499
A(:,:,6,3) =
0.95505 -0.16608 0.12371 0.0067153
-0.16608 0.38639 0.45705 0.02481
0.12371 0.45705 0.65956 -0.01848
0.0067153 0.02481 -0.01848 0.999
A(:,:,1,4) =
0.96774 0.063488 -0.10826 0.12438
0.063488 0.87506 0.21304 -0.24477
-0.10826 0.21304 0.63673 0.41737
0.12438 -0.24477 0.41737 0.52047
A(:,:,2,4) =
0.91488 -0.16388 -0.18495 0.12967
-0.16388 0.6845 -0.35607 0.24964
-0.18495 -0.35607 0.59815 0.28174
0.12967 0.24964 0.28174 0.80247
A(:,:,3,4) =
0.98622 0.043449 -0.066709 0.085142
0.043449 0.86297 0.21038 -0.26852
-0.066709 0.21038 0.67698 0.41227
0.085142 -0.26852 0.41227 0.47382
A(:,:,4,4) =
1 0 0 0
0 1 0 0
0 0 1 0
0 0 0 1
A(:,:,5,4) =
0.73864 0.20112 -0.011394 0.39048
0.20112 0.84524 0.0087678 -0.30047
-0.011394 0.0087678 0.9995 0.017023
0.39048 -0.30047 0.017023 0.41662
A(:,:,6,4) =
0.87322 -0.15647 0.0029936 0.29363
-0.15647 0.80689 0.0036946 0.36238
0.0029936 0.0036946 0.99993 -0.0069332
0.29363 0.36238 -0.0069332 0.31996
A(:,:,1,5) =
0.7542 0.031217 0.42575 0.056052
0.031217 0.99604 -0.054071 -0.0071187
0.42575 -0.054071 0.26255 -0.097088
0.056052 -0.0071187 -0.097088 0.98722
A(:,:,2,5) =
0.95461 0.16812 0.10326 0.066372
0.16812 0.37733 -0.38244 -0.24582
0.10326 -0.38244 0.76511 -0.15098
0.066372 -0.24582 -0.15098 0.90295
A(:,:,3,5) =
0.62859 0.041458 0.47558 0.074661
0.041458 0.99537 -0.053085 -0.0083339
0.47558 -0.053085 0.39105 -0.0956
0.074661 -0.0083339 -0.0956 0.98499
A(:,:,4,5) =
0.73864 0.20112 -0.011394 0.39048
0.20112 0.84524 0.0087678 -0.30047
-0.011394 0.0087678 0.9995 0.017023
0.39048 -0.30047 0.017023 0.41662
A(:,:,5,5) =
1 0 0 0
0 1 0 0
0 0 1 0
0 0 0 1
A(:,:,6,5) =
0.93556 0.24481 -0.0093576 0.016177
0.24481 0.069855 0.035553 -0.061461
-0.0093576 0.035553 0.99864 0.0023492
0.016177 -0.061461 0.0023492 0.99594
A(:,:,1,6) =
0.9818 -0.10286 0.085279 0.0034902
-0.10286 0.41855 0.48208 0.01973
0.085279 0.48208 0.60031 -0.016358
0.0034902 0.01973 -0.016358 0.99933
A(:,:,2,6) =
0.99628 0.012018 0.052874 0.027665
0.012018 0.96117 -0.17085 -0.089393
0.052874 -0.17085 0.24833 -0.39329
0.027665 -0.089393 -0.39329 0.79422
A(:,:,3,6) =
0.95505 -0.16608 0.12371 0.0067153
-0.16608 0.38639 0.45705 0.02481
0.12371 0.45705 0.65956 -0.01848
0.0067153 0.02481 -0.01848 0.999
A(:,:,4,6) =
0.87322 -0.15647 0.0029936 0.29363
-0.15647 0.80689 0.0036946 0.36238
0.0029936 0.0036946 0.99993 -0.0069332
0.29363 0.36238 -0.0069332 0.31996
A(:,:,5,6) =
0.93556 0.24481 -0.0093576 0.016177
0.24481 0.069855 0.035553 -0.061461
-0.0093576 0.035553 0.99864 0.0023492
0.016177 -0.061461 0.0023492 0.99594
A(:,:,6,6) =
1 0 0 0
0 1 0 0
0 0 1 0
0 0 0 1
编辑 2:
添加了相关代码。我一直在暂停代码,最后在 for 循环中得到错误。 (我相信它也在 S
中给出错误,但我一直专注于 D
试图弄清楚。)
mtimesx 来自 here.
n = 4;
M = 6;
P = Normalize(2*rand(n,1,M)-1);
%differences between p_i and p_j
%sum of p_i and p_j
d = Normalize(repmat(permute(P,[1,3,2]),[1,1,M]) - repmat(P,[1,M,1]));
s = Normalize(repmat(permute(P,[1,3,2]),[1,1,M]) + repmat(P,[1,M,1]));
d(isnan(d)) = 0;
%orthogonal projection onto d(:,i,j), i.e. outer product of differences
%orthogonal projection onto s(:,i,j), i.e. outer product of sums
D = mtimesx(permute(d,[1,4,2,3]), permute(d,[4,1,2,3]));
S = mtimesx(permute(s,[1,4,2,3]), permute(s,[4,1,2,3]));
D2 = D;
S2 = S;
%projection onto the complement of d(:,i,j)
%projection onto the complement of s(:,i,j)
D = repmat(eye(n),[1,1,M,M]) - D;
S = repmat(eye(n),[1,1,M,M]) - S;
%total distance to the nearest subspace
PDist = zeros([1,M]);
PDist2 = PDist;
for j = 1:M
for k = 1:M-1
for l = k:M
if j~=k && j~=l
PDist(j) = PDist(j) + min(norm(P(:,1,j) - mtimes(D(:,:,k,l),P(:,1,j))), norm(P(:,1,j) - mtimes(S(:,:,k,l),P(:,1,j))));
PDist2(j) = PDist2(j) + min(norm(D2(:,:,k,1)*P(:,1,j)),norm(S2(:,:,k,1)*P(:,1,j)));
end
end
end
end
PDist-PDist2
Normalize.m
%Normalize
%Accepts an array (of column vectors) and normalizes the columns
function B = Normalize(A)
B = A./repmat(sqrt(sum(A.*A)),size(A,1),1);
end
问题是您在第一个示例和计算 PDist2 的代码中都使用常量 1 而不是变量 l(小写 L)对矩阵进行了索引。
一般来说,最好避免使用看起来彼此相似and/or类似于数字的变量名。
这可以通过使用高亮显示变量和常量使用不同颜色的编辑器来避免(我不知道这在 MATLAB 中是否可行)。实际上,这就是我在您的代码中发现错误的方式。如您所见,当为计算 PDist2 索引 D2 时,数字 1 显示为红色。
设 I
为恒等式,D
为正交投影,p
为向量。
我意识到我的几行代码合并为 (I-(I-D))(p)
,我可以将其简化为 D(p)
。在替换它时,我计算了新方法和旧方法,以仔细检查我是否在计算相同的东西(在我的代码的早期,我有一行是 D = I - D
。你在这里看到的 D
是D
.) 我没有得到相同的答案,并将其追溯到索引 D
.
在这里你可以看到我正在使用调试器并检查 D 的部分并返回错误的数据。
右侧数据资源管理器中的值是我期望的值。有时我从 D(:,:,k,1)
得到了我期望的结果,有时我没有,即使我一个接一个地进行查询也是如此。
那些红色箭头指向的向量应该是相同的。在这些行之间没有任何其他更改或计算,并且当第一行是 运行 时 k = 2
。我关闭了 MATLAB 并重新启动它,每次都遇到同样的问题。 (D 取决于随机输入,但我没有改变种子,所以我在新打开 MATLAB 后第一次 运行 得到相同的东西。D
的计算方式,我 do 期望 D(:,:,1,1)
是单位矩阵。)
这世界到底是怎么回事?感谢任何帮助。
我想知道 MATLAB 是不是故意在搞我。有时当我打开它时,会弹出一个对话框说我需要更新我的学生许可证。我点击了更新按钮,但什么也没有发生,对话框也没有关闭,所以我点击了取消。
编辑:
K>> whos D P
Name Size Bytes Class Attributes
D 4-D 4608 double
P 4x1x6 192 double
K>> size(D)
ans =
4 4 6 6
我一直在研究 A
和 B
,我得到了同样的结果。有时它会正确计算,有时却不会。
K>> B=permute(P,[1,3,2])
B =
0.4155 0.27554 0.52338 0.6991 -0.11346 0.20999
0.53573 -0.83781 0.53182 -0.022364 0.60291 -0.62601
-0.49246 -0.46111 -0.39168 0.45919 0.42377 0.47074
0.54574 0.097595 0.53835 -0.54763 0.66637 0.58516
K>> A=D
A(:,:,1,1) =
1 0 0 0
0 1 0 0
0 0 1 0
0 0 0 1
A(:,:,2,1) =
0.99071 -0.091198 0.0020814 -0.029755
-0.091198 0.10503 0.020426 -0.292
0.0020814 0.020426 0.99953 0.0066643
-0.029755 -0.292 0.0066643 0.90473
A(:,:,3,1) =
0.46769 0.019281 -0.49725 0.036486
0.019281 0.9993 0.018011 -0.0013215
-0.49725 0.018011 0.53551 0.034083
0.036486 -0.0013215 0.034083 0.9975
A(:,:,4,1) =
0.96774 0.063488 -0.10826 0.12438
0.063488 0.87506 0.21304 -0.24477
-0.10826 0.21304 0.63673 0.41737
0.12438 -0.24477 0.41737 0.52047
A(:,:,5,1) =
0.7542 0.031217 0.42575 0.056052
0.031217 0.99604 -0.054071 -0.0071187
0.42575 -0.054071 0.26255 -0.097088
0.056052 -0.0071187 -0.097088 0.98722
A(:,:,6,1) =
0.9818 -0.10286 0.085279 0.0034902
-0.10286 0.41855 0.48208 0.01973
0.085279 0.48208 0.60031 -0.016358
0.0034902 0.01973 -0.016358 0.99933
A(:,:,1,2) =
0.99071 -0.091198 0.0020814 -0.029755
-0.091198 0.10503 0.020426 -0.292
0.0020814 0.020426 0.99953 0.0066643
-0.029755 -0.292 0.0066643 0.90473
A(:,:,2,2) =
1 0 0 0
0 1 0 0
0 0 1 0
0 0 0 1
A(:,:,3,2) =
0.97125 -0.15889 -0.0080537 -0.051131
-0.15889 0.12194 -0.044507 -0.28256
-0.0080537 -0.044507 0.99774 -0.014323
-0.051131 -0.28256 -0.014323 0.90907
A(:,:,4,2) =
0.91488 -0.16388 -0.18495 0.12967
-0.16388 0.6845 -0.35607 0.24964
-0.18495 -0.35607 0.59815 0.28174
0.12967 0.24964 0.28174 0.80247
A(:,:,5,2) =
0.95461 0.16812 0.10326 0.066372
0.16812 0.37733 -0.38244 -0.24582
0.10326 -0.38244 0.76511 -0.15098
0.066372 -0.24582 -0.15098 0.90295
A(:,:,6,2) =
0.99628 0.012018 0.052874 0.027665
0.012018 0.96117 -0.17085 -0.089393
0.052874 -0.17085 0.24833 -0.39329
0.027665 -0.089393 -0.39329 0.79422
A(:,:,1,3) =
0.46769 0.019281 -0.49725 0.036486
0.019281 0.9993 0.018011 -0.0013215
-0.49725 0.018011 0.53551 0.034083
0.036486 -0.0013215 0.034083 0.9975
A(:,:,2,3) =
0.97125 -0.15889 -0.0080537 -0.051131
-0.15889 0.12194 -0.044507 -0.28256
-0.0080537 -0.044507 0.99774 -0.014323
-0.051131 -0.28256 -0.014323 0.90907
A(:,:,3,3) =
1 0 0 0
0 1 0 0
0 0 1 0
0 0 0 1
A(:,:,4,3) =
0.98622 0.043449 -0.066709 0.085142
0.043449 0.86297 0.21038 -0.26852
-0.066709 0.21038 0.67698 0.41227
0.085142 -0.26852 0.41227 0.47382
A(:,:,5,3) =
0.62859 0.041458 0.47558 0.074661
0.041458 0.99537 -0.053085 -0.0083339
0.47558 -0.053085 0.39105 -0.0956
0.074661 -0.0083339 -0.0956 0.98499
A(:,:,6,3) =
0.95505 -0.16608 0.12371 0.0067153
-0.16608 0.38639 0.45705 0.02481
0.12371 0.45705 0.65956 -0.01848
0.0067153 0.02481 -0.01848 0.999
A(:,:,1,4) =
0.96774 0.063488 -0.10826 0.12438
0.063488 0.87506 0.21304 -0.24477
-0.10826 0.21304 0.63673 0.41737
0.12438 -0.24477 0.41737 0.52047
A(:,:,2,4) =
0.91488 -0.16388 -0.18495 0.12967
-0.16388 0.6845 -0.35607 0.24964
-0.18495 -0.35607 0.59815 0.28174
0.12967 0.24964 0.28174 0.80247
A(:,:,3,4) =
0.98622 0.043449 -0.066709 0.085142
0.043449 0.86297 0.21038 -0.26852
-0.066709 0.21038 0.67698 0.41227
0.085142 -0.26852 0.41227 0.47382
A(:,:,4,4) =
1 0 0 0
0 1 0 0
0 0 1 0
0 0 0 1
A(:,:,5,4) =
0.73864 0.20112 -0.011394 0.39048
0.20112 0.84524 0.0087678 -0.30047
-0.011394 0.0087678 0.9995 0.017023
0.39048 -0.30047 0.017023 0.41662
A(:,:,6,4) =
0.87322 -0.15647 0.0029936 0.29363
-0.15647 0.80689 0.0036946 0.36238
0.0029936 0.0036946 0.99993 -0.0069332
0.29363 0.36238 -0.0069332 0.31996
A(:,:,1,5) =
0.7542 0.031217 0.42575 0.056052
0.031217 0.99604 -0.054071 -0.0071187
0.42575 -0.054071 0.26255 -0.097088
0.056052 -0.0071187 -0.097088 0.98722
A(:,:,2,5) =
0.95461 0.16812 0.10326 0.066372
0.16812 0.37733 -0.38244 -0.24582
0.10326 -0.38244 0.76511 -0.15098
0.066372 -0.24582 -0.15098 0.90295
A(:,:,3,5) =
0.62859 0.041458 0.47558 0.074661
0.041458 0.99537 -0.053085 -0.0083339
0.47558 -0.053085 0.39105 -0.0956
0.074661 -0.0083339 -0.0956 0.98499
A(:,:,4,5) =
0.73864 0.20112 -0.011394 0.39048
0.20112 0.84524 0.0087678 -0.30047
-0.011394 0.0087678 0.9995 0.017023
0.39048 -0.30047 0.017023 0.41662
A(:,:,5,5) =
1 0 0 0
0 1 0 0
0 0 1 0
0 0 0 1
A(:,:,6,5) =
0.93556 0.24481 -0.0093576 0.016177
0.24481 0.069855 0.035553 -0.061461
-0.0093576 0.035553 0.99864 0.0023492
0.016177 -0.061461 0.0023492 0.99594
A(:,:,1,6) =
0.9818 -0.10286 0.085279 0.0034902
-0.10286 0.41855 0.48208 0.01973
0.085279 0.48208 0.60031 -0.016358
0.0034902 0.01973 -0.016358 0.99933
A(:,:,2,6) =
0.99628 0.012018 0.052874 0.027665
0.012018 0.96117 -0.17085 -0.089393
0.052874 -0.17085 0.24833 -0.39329
0.027665 -0.089393 -0.39329 0.79422
A(:,:,3,6) =
0.95505 -0.16608 0.12371 0.0067153
-0.16608 0.38639 0.45705 0.02481
0.12371 0.45705 0.65956 -0.01848
0.0067153 0.02481 -0.01848 0.999
A(:,:,4,6) =
0.87322 -0.15647 0.0029936 0.29363
-0.15647 0.80689 0.0036946 0.36238
0.0029936 0.0036946 0.99993 -0.0069332
0.29363 0.36238 -0.0069332 0.31996
A(:,:,5,6) =
0.93556 0.24481 -0.0093576 0.016177
0.24481 0.069855 0.035553 -0.061461
-0.0093576 0.035553 0.99864 0.0023492
0.016177 -0.061461 0.0023492 0.99594
A(:,:,6,6) =
1 0 0 0
0 1 0 0
0 0 1 0
0 0 0 1
编辑 2:
添加了相关代码。我一直在暂停代码,最后在 for 循环中得到错误。 (我相信它也在 S
中给出错误,但我一直专注于 D
试图弄清楚。)
mtimesx 来自 here.
n = 4;
M = 6;
P = Normalize(2*rand(n,1,M)-1);
%differences between p_i and p_j
%sum of p_i and p_j
d = Normalize(repmat(permute(P,[1,3,2]),[1,1,M]) - repmat(P,[1,M,1]));
s = Normalize(repmat(permute(P,[1,3,2]),[1,1,M]) + repmat(P,[1,M,1]));
d(isnan(d)) = 0;
%orthogonal projection onto d(:,i,j), i.e. outer product of differences
%orthogonal projection onto s(:,i,j), i.e. outer product of sums
D = mtimesx(permute(d,[1,4,2,3]), permute(d,[4,1,2,3]));
S = mtimesx(permute(s,[1,4,2,3]), permute(s,[4,1,2,3]));
D2 = D;
S2 = S;
%projection onto the complement of d(:,i,j)
%projection onto the complement of s(:,i,j)
D = repmat(eye(n),[1,1,M,M]) - D;
S = repmat(eye(n),[1,1,M,M]) - S;
%total distance to the nearest subspace
PDist = zeros([1,M]);
PDist2 = PDist;
for j = 1:M
for k = 1:M-1
for l = k:M
if j~=k && j~=l
PDist(j) = PDist(j) + min(norm(P(:,1,j) - mtimes(D(:,:,k,l),P(:,1,j))), norm(P(:,1,j) - mtimes(S(:,:,k,l),P(:,1,j))));
PDist2(j) = PDist2(j) + min(norm(D2(:,:,k,1)*P(:,1,j)),norm(S2(:,:,k,1)*P(:,1,j)));
end
end
end
end
PDist-PDist2
Normalize.m
%Normalize
%Accepts an array (of column vectors) and normalizes the columns
function B = Normalize(A)
B = A./repmat(sqrt(sum(A.*A)),size(A,1),1);
end
问题是您在第一个示例和计算 PDist2 的代码中都使用常量 1 而不是变量 l(小写 L)对矩阵进行了索引。
一般来说,最好避免使用看起来彼此相似and/or类似于数字的变量名。
这可以通过使用高亮显示变量和常量使用不同颜色的编辑器来避免(我不知道这在 MATLAB 中是否可行)。实际上,这就是我在您的代码中发现错误的方式。如您所见,当为计算 PDist2 索引 D2 时,数字 1 显示为红色。