Classification problem
in [Matlab]
|
Prev: Report Generator IO Documentation
Next: Substring????
From: Karina on 2 Mar 2010 14:29 Hi, I need help with the classification of data points. I have the following array: StaveRows: 96 100 103 107 111 140 144 147 148 151 155 205 206 209 213 216 217 220 250 253 254 257 261 264 265 315 318 319 323 326 330 359 363 366 367 370 374 424 425 428 432 435 436 439 469 472 476 479 480 483 The array represents the coordinates of the Staves in the music score. The problem is that some error points present.So I plotted them into graph (the points are - y axe points) and they visually form 8 groups(8 staves) on the graph. I though, maybe I can classify them into groups and then find the errors. What kind of classification is better? I tried to use k means clustering, it didnt work. I think it can work for only 1 iteration. Is there a way to make k means clustering only for 1 iteration?
From: Greg Heath on 3 Mar 2010 00:29 On Mar 2, 2:29 pm, "Karina " <miss.v...(a)gmail.com> wrote: > Hi, > I need help with the classification of data points. > I have the following array: > StaveRows: > 96 > 100 > 103 > 107 > 111 > 140 > 144 > 147 > 148 > 151 > 155 > 205 > 206 > 209 > 213 > 216 > 217 > 220 > 250 > 253 > 254 > 257 > 261 > 264 > 265 > 315 > 318 > 319 > 323 > 326 > 330 > 359 > 363 > 366 > 367 > 370 > 374 > 424 > 425 > 428 > 432 > 435 > 436 > 439 > 469 > 472 > 476 > 479 > 480 > 483 > The array represents the coordinates of the Staves in the music score. The problem is that some error points present.So I plotted them into graph (the points are - y axe points) and they visually form 8 groups(8 staves) on the graph. I though, maybe I can classify them into groups and then find the errors. > What kind of classification is better? I tried to use k means clustering, it didnt work. I think it can work for only 1 iteration. Is there a way to make k means clustering only for 1 iteration? help diff Hope this helps. Greg
From: Peter Perkins on 3 Mar 2010 07:37 On 3/2/2010 2:29 PM, Karina wrote: > The array represents the coordinates of the Staves in the music score. > The problem is that some error points present.So I plotted them into > graph (the points are - y axe points) and they visually form 8 groups(8 > staves) on the graph. I though, maybe I can classify them into groups > and then find the errors. What kind of classification is better? I tried > to use k means clustering, it didnt work. I think it can work for only 1 > iteration. Is there a way to make k means clustering only for 1 iteration? KMEANS does indeed work on these data. As the doc says, <http://www.mathworks.com/access/helpdesk/help/toolbox/stats/bq_679x-18.html> "Like many other types of numerical minimizations, the solution that kmeans reaches often depends on the starting points. ... However, you can use the optional 'replicates' parameter to overcome that problem." So for example, cidx = kmeans(x,8,'Replicates',10); plot(cidx,x,'.');
From: Karina on 4 Mar 2010 23:08 Peter Perkins <Peter.Perkins(a)MathRemoveThisWorks.com> wrote in message <hmll6d$i25$1(a)fred.mathworks.com>... > On 3/2/2010 2:29 PM, Karina wrote: > > The array represents the coordinates of the Staves in the music score. > > The problem is that some error points present.So I plotted them into > > graph (the points are - y axe points) and they visually form 8 groups(8 > > staves) on the graph. I though, maybe I can classify them into groups > > and then find the errors. What kind of classification is better? I tried > > to use k means clustering, it didnt work. I think it can work for only 1 > > iteration. Is there a way to make k means clustering only for 1 iteration? > > KMEANS does indeed work on these data. As the doc says, > > <http://www.mathworks.com/access/helpdesk/help/toolbox/stats/bq_679x-18.html> > > "Like many other types of numerical minimizations, the solution that kmeans reaches often depends on the starting points. ... However, you can use the optional 'replicates' parameter to overcome that problem." > > So for example, > It works! Thanx a lot! > cidx = kmeans(x,8,'Replicates',10); > plot(cidx,x,'.');
From: Peter Perkins on 5 Mar 2010 13:30
On 3/4/2010 11:08 PM, Karina wrote: > Peter Perkins <Peter.Perkins(a)MathRemoveThisWorks.com> wrote in message >> cidx = kmeans(x,8,'Replicates',10); >> plot(cidx,x,'.'); I should probably point out that since you already know that the centroids will be at something close to 1:8, you could also have provided that as an explicit starting point, and used only one replicate. But I don't know how general your code needs to be. |