Classify data points
in [Matlab]
|
From: Diego Lass on 24 May 2010 19:15 Hi there, I want to classify data points based on its distance from two values, example x = [1 2 3 4 5]; y = 2.3; z = 4.5; I want all the points less than y classified as 0, greater than z classified as 1. And if between y and z, classify according to its distance to y and z (take the shorter distance) Result for above is result = [0 0 0 1 1] How can I do this as efficient as possible? Thanks Diego
From: Walter Roberson on 24 May 2010 19:42 Diego Lass wrote: > I want to classify data points based on its distance from two values, > example > > x = [1 2 3 4 5]; > y = 2.3; > z = 4.5; > > I want all the points less than y classified as 0, greater than z > classified as 1. And if between y and z, classify according to its > distance to y and z (take the shorter distance) > Result for above is > result = [0 0 0 1 1] > > How can I do this as efficient as possible? What if something is exactly half way between the y and z? The below will count the exact half way point as rounding to z. [counts, result] = histc(x, [(y+z)/2 inf]); Unless, that is, one of the x can actually be inf, as that would return a bin of 2.
From: ImageAnalyst on 24 May 2010 20:35 Diego: A little verbose and explicit perhaps but that's just to help you understand it better. But it works, and without any loops. % Sample data: x = [1 2 3 4 5]; y = 2.3; z = 4.5; % Prepare an output matrix. output = x; % Assign the easy ones. output(x < y) = 0; output(x > z) = 1; indexesAlreadyAssigned = (x < y) | (x > z) % Find out the distance from x to y. differenceFromY = abs(x-y) % Find out the distance from x to z. differenceFromZ = abs(x-z) % Find out where distance to y is less than distance to z. takeYif = differenceFromY < differenceFromZ % Find out where distance to z is less than distance to y. takeZif = differenceFromZ < differenceFromY % Zero out indexes that we've already used. takeYif(indexesAlreadyAssigned) = 0 takeZif(indexesAlreadyAssigned) = 0 % For the "in between" values, take the one that's closest. output(takeYif) = 0; output(takeZif) = 1; % Display the output. output indexesAlreadyAssigned = 1 1 0 0 1 differenceFromY = 1.3000 0.3000 0.7000 1.7000 2.7000 differenceFromZ = 3.5000 2.5000 1.5000 0.5000 0.5000 takeYif = 1 1 1 0 0 takeZif = 0 0 0 1 1 takeYif = 0 0 1 0 0 takeZif = 0 0 0 1 0 output = 0 0 0 1 1
From: Diego Lass on 24 May 2010 22:10 Thanks! Diego ImageAnalyst <imageanalyst(a)mailinator.com> wrote in message <92f32a40-4a1c-4cf5-9a16-ec5a574fdb39(a)c22g2000vbb.googlegroups.com>... > Diego: > A little verbose and explicit perhaps but that's just to help you > understand it better. But it works, and without any loops. > > % Sample data: > x = [1 2 3 4 5]; > y = 2.3; > z = 4.5; > > % Prepare an output matrix. > output = x; > % Assign the easy ones. > output(x < y) = 0; > output(x > z) = 1; > indexesAlreadyAssigned = (x < y) | (x > z) > % Find out the distance from x to y. > differenceFromY = abs(x-y) > % Find out the distance from x to z. > differenceFromZ = abs(x-z) > % Find out where distance to y is less than distance to z. > takeYif = differenceFromY < differenceFromZ > % Find out where distance to z is less than distance to y. > takeZif = differenceFromZ < differenceFromY > % Zero out indexes that we've already used. > takeYif(indexesAlreadyAssigned) = 0 > takeZif(indexesAlreadyAssigned) = 0 > % For the "in between" values, take the one that's closest. > output(takeYif) = 0; > output(takeZif) = 1; > % Display the output. > output > > indexesAlreadyAssigned = > 1 1 0 0 1 > > differenceFromY = > 1.3000 0.3000 0.7000 1.7000 2.7000 > > differenceFromZ = > 3.5000 2.5000 1.5000 0.5000 0.5000 > > takeYif = > 1 1 1 0 0 > > takeZif = > 0 0 0 1 1 > > takeYif = 0 0 1 0 0 > > takeZif = > 0 0 0 1 0 > > output = > 0 0 0 1 1 >
From: Greg Heath on 25 May 2010 08:41 On May 24, 7:15 pm, "Diego Lass" <dlISC...(a)gmail.com> wrote: > Hi there, > I want to classify data points based on its distance from two values, example > > x = [1 2 3 4 5]; > y = 2.3; > z = 4.5; > > I want all the points less than y classified as 0, greater than z classified as 1. And if between y and z, classify according to its distance to y and z (take the shorter distance) > Result for above is > result = [0 0 0 1 1] > > How can I do this as efficient as possible? Thanks > Diego compare abs(x-y) with abs(x-z) Hope this helps. Greg
|
Pages: 1 Prev: output numbers from a Simulink model on a window as a number Next: combining vectors |