Let I be the identity, D an orthonormal projection, and p a vector.
I realized that several of my lines of code combined to be (I-(I-D))(p) and I could just simplify it to D(p). In replacing it, I computed the new method along-side the old to double check I was computing the same thing (Earlier in my code I had a line that was D = I - D. The D you see here is that D.) I wasn't getting the same answer, and traced it to an error in indexing D.
Here you can see I'm using the debugger and checking portions of D and getting the wrong data returned.
The values in the data explorer on the right are what I'd expect them to be. Sometimes I get what I'd expect from D(:,:,k,1), and sometimes I don't, even when I make the queries right after each other.
The vectors those red arrows are pointing to should be the same. Nothing else changed or was computed between those lines, and k = 2 when the first line was run. I've closed MATLAB and restarted it and get the same issue every time. (D depends on random input, but I'm not altering the seed, so I get the same thing every first run after newly opening MATLAB. The way D is computed, I do expect D(:,:,1,1) to be the identity matrix.)
What in the world is going on? Any help is appreciated.
I have wondered if MATLAB is messing with me on purpose. Sometimes when I open it, a pop-up dialog box says I need to update my student license. I click the update button, but nothing ever happens and the dialog box never closes, so I click cancel.
Edit:
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
I've been playing around with A and B a bit, and I get the same thing. Sometimes it computes correctly and sometimes it doesn't.
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
Edit 2:
Added relevant code. I've been pausing the code and getting the errors inside the for loops at the end. (I believe it's also giving errors in S, but I've been focusing on D trying to figure it out.)
mtimesx is from 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
The problem is that you indexed the matrices using the constant 1 instead of the variable l (lowercase L), both in the first example and in the code for computing PDist2.
In general it is good to avoid using variable names that look similar to each other and/or similar to numbers.
This can be avoided by using an editor that highlights uses different colors for variables and constants (I don't know if this is possible in MATLAB). In fact, this is how I found the error in your code. As you can see, when indexing D2 for the computation of PDist2 the number 1 is colored red.
My project is "optical flow estimation for flame detection in videos" In that while extracting feature values, I can only retain the last intensity value of the frame.
Here is my code
function [Iy, Ix, It] = grad3D(imNew,bFineScale,bInitialize)
persistent siz gx gg imPrev;
if nargin>2 && bInitialize
[gx, gg]= makeFilters();
if bFineScale
siz = size(imNew);
imPrev= single(imNew);
else% if ~bFineScale
siz = floor(size(imNew)/2);
%initialize imPrev to half the size
imPrev = imresizeNN(single(imNew),siz);
end
end
if ~bFineScale
imNew = imresizeNN(conv2(single(imNew),gg,'same'),siz);
else imNew = single(imNew);
end
Ix = conv2(gg,gx,imNew + imPrev,'same');
Iy = conv2(gx,gg,imNew + imPrev,'same');
It = conv2(gg,gg,imNew - imPrev,'same'); %L3
% finally, store away the current image for use on next frame
imPrev = imNew;
testfeature = mean(imPrev);
save testfeature testfeature
[gx, gg]= makeFilters() x = (-1:1);
gg = single(gaussgen(0.67,3));
gx = single(-x.*gg*3);
In the highlighted coding(testfeature=mean(imprev)). I can be able to get only the intensity value of last frame extracted...But i need the values for all the extracted frames. I need the value to be stored row wise in a matrix file.
I wrote a script that returns several text boxes in a figure. The text boxes are moveable (I can drag and drop them), and their positions are predetermined by the data in an input matrix (the data from the input matrix is applied to the respective positions of the boxes by nested for loop). I want to create a matrix which is initially a copy of the input matrix, but is UPDATED as I change the positions of the boxes by dragging them around. How would I update their positions? Here's the entire script
function drag_drop=drag_drop(tsinput,infoinput)
[x,~]=size(tsinput);
dragging = [];
orPos = [];
fig = figure('Name','Docker Tool','WindowButtonUpFcn',#dropObject,...
'units','centimeters','WindowButtonMotionFcn',#moveObject,...
'OuterPosition',[0 0 25 30]);
% Setting variables to zero for the loop
plat_qty=0;
time_qty=0;
k=0;
a=0;
% Start loop
z=1:2
for idx=1:x
if tsinput(idx,4)==1
color='red';
else
color='blue';
end
a=tsinput(idx,z);
b=a/100;
c=floor(b); % hours
d=c*100;
e=a-d; % minutes
time=c*60+e; % time quantity to be used in 'position'
time_qty=time/15;
plat_qty=tsinput(idx,3)*2;
box=annotation('textbox','units','centimeters','position',...
[time_qty plat_qty 1.5 1.5],'String',infoinput(idx,z),...
'ButtonDownFcn',#dragObject,'BackgroundColor',color);
% need to new=get(box,'Position'), fill out matrix OUT of loop
end
fillmenu=uicontextmenu;
hcb1 = 'set(gco, ''BackgroundColor'', ''red'')';
hcb2 = 'set(gco, ''BackgroundColor'', ''blue'')';
item1 = uimenu(fillmenu, 'Label', 'Train Full', 'Callback', hcb1);
item2 = uimenu(fillmenu, 'Label', 'Train Empty', 'Callback', hcb2);
hbox=findall(fig,'Type','hggroup');
for jdx=1:x
set(hbox(jdx),'uicontextmenu',fillmenu);
end
end
new_arr=tsinput;
function dragObject(hObject,eventdata)
dragging = hObject;
orPos = get(gcf,'CurrentPoint');
end
function dropObject(hObject,eventdata,box)
if ~isempty(dragging)
newPos = get(gcf,'CurrentPoint');
posDiff = newPos - orPos;
set(dragging,'Position',get(dragging,'Position') + ...
[posDiff(1:2) 0 0]);
dragging = [];
end
end
function moveObject(hObject,eventdata)
if ~isempty(dragging)
newPos = get(gcf,'CurrentPoint');
posDiff = newPos - orPos;
orPos = newPos;
set(dragging,'Position',get(dragging,'Position') + [posDiff(1:2) 0 0]);
end
end
end
% Testing purpose input matrices:
% tsinput=[0345 0405 1 1 ; 0230 0300 2 0; 0540 0635 3 1; 0745 0800 4 1]
% infoinput={'AJ35 NOT' 'KL21 MAN' 'XPRES'; 'ZW31 MAN' 'KM37 NEW' 'VISTA';
% 'BC38 BIR' 'QU54 LON' 'XPRES'; 'XZ89 LEC' 'DE34 MSF' 'DERP'}
If I understand you correctly (and please post some code if I'm not), then all you need is indeed a set/get combination.
If boxHandle is a handle to the text-box object, then you get its current position by:
pos = get (boxHandle, 'position')
where pos is the output array of [x, y, width, height].
In order to set to a new position, you use:
set (boxHandle, 'position', newPos)
where newPos is the array of desired position (with the same structure as pos).
EDIT
Regarding to updating your matrix, since you have the handle of the object you move, you actually DO have access to the specific text box.
When you create each text box, set a property called 'UserData' with the associated indices of tsinput used for that box. In your nested for loop add this
set (box, 'UserData', [idx, z]);
after the box is created, and in your moveObject callback get the data by
udata = get(dragging,'UserData');
Then udata contains the indices of the elements you want to update.