Thread Subject: Get line index on matrix

Hello,

I'm trying to get the line index of a nx3 matrix for a given vector (1x3), like

0.1 0.3 1.6
0.2 0.7 1.2
1.3 0.1 7.3
(...)

searching for vector = 0.2 0.7 1.2 should return index = 2.

i've tried find() and a script which compares

    if(vector(i,:) == points(i,:))
        index = i;

inside a for() loop (but it says that the condition is not valid, on debug)

I would appreciate your help.

Best regards

"Joe Nunes" <vazdepaivanunes@gmail.com> wrote in message <hcudtq$phk$1@fred.mathworks.com>...
> Hello,
>
> I'm trying to get the line index of a nx3 matrix for a given vector (1x3), like
>
> 0.1 0.3 1.6
> 0.2 0.7 1.2
> 1.3 0.1 7.3
> (...)
>
> searching for vector = 0.2 0.7 1.2 should return index = 2.
>
> i've tried find() and a script which compares
>
> if(vector(i,:) == points(i,:))
> index = i;
>
> inside a for() loop (but it says that the condition is not valid, on debug)
>
> I would appreciate your help.
>
> Best regards

Hi Joe, will the following work for you?

A=[0.1 0.3 1.6
0.2 0.7 1.2
1.3 0.1 7.3];
B= [0.2 0.7 1.2];
[C,IA] = intersect(A,B,'rows');
% IA is equal to 2
IA

Wayne

"Wayne King" <wmkingty@gmail.com> wrote in message <hcuetm$6t$1@fred.mathworks.com>...
> "Joe Nunes" <vazdepaivanunes@gmail.com> wrote in message <hcudtq$phk$1@fred.mathworks.com>...
> > Hello,
> >
> > I'm trying to get the line index of a nx3 matrix for a given vector (1x3), like
> >
> > 0.1 0.3 1.6
> > 0.2 0.7 1.2
> > 1.3 0.1 7.3
> > (...)
> >
> > searching for vector = 0.2 0.7 1.2 should return index = 2.
> >
> > i've tried find() and a script which compares
> >
> > if(vector(i,:) == points(i,:))
> > index = i;
> >
> > inside a for() loop (but it says that the condition is not valid, on debug)
> >
> > I would appreciate your help.
> >
> > Best regards
>
> Hi Joe, will the following work for you?
>
> A=[0.1 0.3 1.6
> 0.2 0.7 1.2
> 1.3 0.1 7.3];
> B= [0.2 0.7 1.2];
> [C,IA] = intersect(A,B,'rows');
> % IA is equal to 2
> IA
>
> Wayne

Yes, i've tried, but IA returns an Empty matrix: 0-by-1 .. Since my set of points is 1312x3 and the points to find are in 17x3, i've tried
[C,IA] = intersect(points,keypoints(2,:),'rows')

but it keeps returning empty matrix. Is this because of rounding or something ? i don't understand. This points from 17x3 matrix were taken from the original set.

"Wayne King" <wmkingty@gmail.com> wrote in message <hcuetm$6t$1@fred.mathworks.com>...
> "Joe Nunes" <vazdepaivanunes@gmail.com> wrote in message <hcudtq$phk$1@fred.mathworks.com>...
> > Hello,
> >
> > I'm trying to get the line index of a nx3 matrix for a given vector (1x3), like
> >
> > 0.1 0.3 1.6
> > 0.2 0.7 1.2
> > 1.3 0.1 7.3
> > (...)
> >
> > searching for vector = 0.2 0.7 1.2 should return index = 2.
> >
> > i've tried find() and a script which compares
> >
> > if(vector(i,:) == points(i,:))
> > index = i;
> >
> > inside a for() loop (but it says that the condition is not valid, on debug)
> >
> > I would appreciate your help.
> >
> > Best regards
>
> Hi Joe, will the following work for you?
>
> A=[0.1 0.3 1.6
> 0.2 0.7 1.2
> 1.3 0.1 7.3];
> B= [0.2 0.7 1.2];
> [C,IA] = intersect(A,B,'rows');
> % IA is equal to 2
> IA
>
> Wayne

or take a look at ISMEMBER or even more directly BSXFUN (in combination with ALL)

A=[0.1 0.3 1.6 ; 0.2 0.7 1.2 ; 1.3 0.1 7.3]; B= [0.2 0.7 1.2];
t1 = bsxfun(@eq,A,B)
t2 = all(t1,2)
r = find(t2)

By the way, be quite careful when comparing floating points:

a = 0.1+0.2+0.3, b = 0.6, a==b

hth
Jos

"Joe Nunes" <vazdepaivanunes@gmail.com> wrote in message <hcufeh$298$1@fred.mathworks.com>...
> "Wayne King" <wmkingty@gmail.com> wrote in message <hcuetm$6t$1@fred.mathworks.com>...
> > "Joe Nunes" <vazdepaivanunes@gmail.com> wrote in message <hcudtq$phk$1@fred.mathworks.com>...
> > > Hello,
> > >
> > > I'm trying to get the line index of a nx3 matrix for a given vector (1x3), like
> > >
> > > 0.1 0.3 1.6
> > > 0.2 0.7 1.2
> > > 1.3 0.1 7.3
> > > (...)
> > >
> > > searching for vector = 0.2 0.7 1.2 should return index = 2.
> > >
> > > i've tried find() and a script which compares
> > >
> > > if(vector(i,:) == points(i,:))
> > > index = i;
> > >
> > > inside a for() loop (but it says that the condition is not valid, on debug)
> > >
> > > I would appreciate your help.
> > >
> > > Best regards
> >
> > Hi Joe, will the following work for you?
> >
> > A=[0.1 0.3 1.6
> > 0.2 0.7 1.2
> > 1.3 0.1 7.3];
> > B= [0.2 0.7 1.2];
> > [C,IA] = intersect(A,B,'rows');
> > % IA is equal to 2
> > IA
> >
> > Wayne
>
> Yes, i've tried, but IA returns an Empty matrix: 0-by-1 .. Since my set of points is 1312x3 and the points to find are in 17x3, i've tried
> [C,IA] = intersect(points,keypoints(2,:),'rows')
>
> but it keeps returning empty matrix. Is this because of rounding or something ? i don't understand. This points from 17x3 matrix were taken from the original set.

Hi, Maybe.. are you doing this inside of an M-file? If you do it in the command window (outside of an M-file) does it work? How about if you force them both to be doubles?

s = RandStream('mt19937ar');
RandStream.setDefaultStream(s);
A = randn(1000,3);
 B=A(4,:);
 [C,IA] = intersect(double(A),double(B),'rows');
% IA is 4
IA

wayne

i've tried forcing to double. that didn't work either. i don't have the bsxfun function, i've looked to ismember though .. =(

"Wayne King" <wmkingty@gmail.com> wrote in message <hcugch$fv$1@fred.mathworks.com>...
> "Joe Nunes" <vazdepaivanunes@gmail.com> wrote in message <hcufeh$298$1@fred.mathworks.com>...
> > "Wayne King" <wmkingty@gmail.com> wrote in message <hcuetm$6t$1@fred.mathworks.com>...
> > > "Joe Nunes" <vazdepaivanunes@gmail.com> wrote in message <hcudtq$phk$1@fred.mathworks.com>...
> > > > Hello,
> > > >
> > > > I'm trying to get the line index of a nx3 matrix for a given vector (1x3), like
> > > >
> > > > 0.1 0.3 1.6
> > > > 0.2 0.7 1.2
> > > > 1.3 0.1 7.3
> > > > (...)
> > > >
> > > > searching for vector = 0.2 0.7 1.2 should return index = 2.
> > > >
> > > > i've tried find() and a script which compares
> > > >
> > > > if(vector(i,:) == points(i,:))
> > > > index = i;
> > > >
> > > > inside a for() loop (but it says that the condition is not valid, on debug)
> > > >
> > > > I would appreciate your help.
> > > >
> > > > Best regards
> > >
> > > Hi Joe, will the following work for you?
> > >
> > > A=[0.1 0.3 1.6
> > > 0.2 0.7 1.2
> > > 1.3 0.1 7.3];
> > > B= [0.2 0.7 1.2];
> > > [C,IA] = intersect(A,B,'rows');
> > > % IA is equal to 2
> > > IA
> > >
> > > Wayne
> >
> > Yes, i've tried, but IA returns an Empty matrix: 0-by-1 .. Since my set of points is 1312x3 and the points to find are in 17x3, i've tried
> > [C,IA] = intersect(points,keypoints(2,:),'rows')
> >
> > but it keeps returning empty matrix. Is this because of rounding or something ? i don't understand. This points from 17x3 matrix were taken from the original set.
>
> Hi, Maybe.. are you doing this inside of an M-file? If you do it in the command window (outside of an M-file) does it work? How about if you force them both to be doubles?
>
> s = RandStream('mt19937ar');
> RandStream.setDefaultStream(s);
> A = randn(1000,3);
> B=A(4,:);
> [C,IA] = intersect(double(A),double(B),'rows');
> % IA is 4
> IA
>
> wayne

Are you aware of section 6.1 of the FAQ, which relates to comparing
floating point numbers for exact equality?
http://matlabwiki.mathworks.com/MATLAB_FAQ

"Joe Nunes" <vazdepaivanunes@gmail.com> wrote in message <hcuph2$2mp$1@fred.mathworks.com>...
> i've tried forcing to double. that didn't work either. i don't have the bsxfun function, i've looked to ismember though .. =(

So, what did you look at? Did you try anything?

Jos

Joe Nunes wrote:
...
> Yes, i've tried, but IA returns an Empty matrix: 0-by-1 .. Since my
> set of points is 1312x3 and the points to find are in 17x3, i've
> tried [C,IA] = intersect(points,keypoints(2,:),'rows')
>
> but it keeps returning empty matrix. Is this because of rounding or
> something ? i don't understand. This points from 17x3 matrix were
> taken from the original set.

Well, could well be/probably is--all it takes is the last bit of the
mantissa to be different for the comparison to fail on equality.

You might look at comparing the individual values and outputting that
difference/equality result to see where the actual failure is occurring.
  If you're doing any computation or i/o or other operations between
there are chances for rounding to occur.

But, in general, for robust coding you'll need to do some sort of
tolerancing on the interval for selection of floating point values or
some case will inevitably come back to bite even if you figure out a
workaround for this specific set of values.

Somebody else noted the FAQ.

--

Replace ISMEMBER call with ISMEMBERF on FEX:

http://www.mathworks.com/matlabcentral/fileexchange/23294-ismemberf

Bruno

"Bruno Luong" <b.luong@fogale.findmycountry> wrote in message <hd0j46$n1i$1@fred.mathworks.com>...
> Replace ISMEMBER call with ISMEMBERF on FEX:
>
> http://www.mathworks.com/matlabcentral/fileexchange/23294-ismemberf
>
> Bruno

Example:

>> vector=[1 3 16;
                  2 7 12;
                  13 1 73]*(1/10)

vector =

    0.1000 0.3000 1.6000
    0.2000 0.7000 1.2000
    1.3000 0.1000 7.3000

>> [t idx]=ismember([0.2 0.7 1.2],vector,'rows') % <- exact comparison

t =

     0


idx =

     0

>> [t idx]=ismemberf([0.2 0.7 1.2],vector,'rows')

t =

     1


idx =

     2

% Bruno

It worked!!!!! thank you very much!

Best Regards

"Bruno Luong" <b.luong@fogale.findmycountry> wrote in message <hd0jqm$67p$1@fred.mathworks.com>...
> "Bruno Luong" <b.luong@fogale.findmycountry> wrote in message <hd0j46$n1i$1@fred.mathworks.com>...
> > Replace ISMEMBER call with ISMEMBERF on FEX:
> >
> > http://www.mathworks.com/matlabcentral/fileexchange/23294-ismemberf
> >
> > Bruno
>
> Example:
>
> >> vector=[1 3 16;
> 2 7 12;
> 13 1 73]*(1/10)
>
> vector =
>
> 0.1000 0.3000 1.6000
> 0.2000 0.7000 1.2000
> 1.3000 0.1000 7.3000
>
> >> [t idx]=ismember([0.2 0.7 1.2],vector,'rows') % <- exact comparison
>
> t =
>
> 0
>
>
> idx =
>
> 0
>
> >> [t idx]=ismemberf([0.2 0.7 1.2],vector,'rows')
>
> t =
>
> 1
>
>
> idx =
>
> 2
>
> % Bruno