iFit/kmeans

PURPOSE

SYNOPSIS

DESCRIPTION

 [b,c] = kmeans(X, k) : k-means clustering of iData object

   @iData/kmeans function to partition the object X into k classes.

   b = kmeans(X,k) partitions the Signal in the iData object X into k clusters.
     The resulting object Signal contains numbers from 1 to 'k' which are indices
     of segments/partitions.
     When no cluster can be found, the result is empty.
   b = kmeans(X)   assumes k=2 partitions

   [b,c] = kmeans(X,k) also returns the centroid of the clusters/partitions/segments.

   b = kmeans(X, k, 'otsu')
     uses the Otsu method, for 2D data sets
   b = kmeans(X, k, 'kmeans++')
     uses the K-means++ method, for 2D data sets

 input:  X: object or array (iData)
         k: number of partitions wanted (integer, default is 2)
         method: '' for default, 'otsu' for Otsu method (only for 2D images).
 output: b: object or array with partition indices (iData)
         c: centroid locations of clusters
 ex:     a=iData(peaks); b=kmeans(a);

 See: https://en.wikipedia.org/wiki/K-means_clustering
 See: https://en.wikipedia.org/wiki/K-means%2B%2B
 See: https://en.wikipedia.org/wiki/Otsu%27s_method

 References:
   Otsu N., A threshold selection method from gray-level histogram, IEEE Trans. Syst. Man Cybern. 9:62-66;1979
   [1] J. B. MacQueen, "Some Methods for Classification and Analysis of 
       MultiVariate Observations", in Proc. of the fifth Berkeley
       Symposium on Mathematical Statistics and Probability, L. M. L. Cam
       and J. Neyman, eds., vol. 1, UC Press, 1967, pp. 281-297.
   [2] D. Arthur and S. Vassilvitskii, "k-means++: The Advantages of
       Careful Seeding", Technical Report 2006-13, Stanford InfoLab, 2006.

 Version: Nov. 27, 2018
 See also iData, iData/uminus, iData/abs, iData/real, iData/imag, iData/uplus

CROSS-REFERENCE INFORMATION

• abs b = abs(s) : absolute value of iData object
• cell [content,fields]=cell(s) : convert iData objects into cells
• copyobj b = copyobj(s) : makes a copy of iData object
• cumsum s = cumsum(a,dim) : computes the cumulative sum of iData objects elements
• end b = end(s,index,n) : end value for iData objects
• find d = find(s) : find iData signal non zeros values
• fliplr b = fliplr(s) : Flip object in left/right direction
• hist hist(a, axis1, axis2, ...) computes the accumulated histogram from event data sets
• iData
• ind2sub ind2sub(s,index) : get indexed element in an iData array
• isempty isempty(s) : true for empty iData object
• isfinite b = isfinite(s) : True for finite iData object elements
• isfloat b = isfloat(s) : True for float iData object elements
• isnan b = isnan(s) : True for NaN iData object elements
• label b = label(s, alias, label) : Change iData label for a given alias/axis
• linspace v = linspace(a,b,n) : creates a linearly spaced vector of objects
• max [m,id] = max(a,b, dim) : computes the maximum value of iData object(s)
• min [m,id] = min(a,b, dim) : computes the maximum value of iData object(s)
• minus c = minus(a,b) : computes the difference of iData objects
• ndims d = ndims(s) : get the dimensionality of iData object
• ones s = ones(s,N,M,P,...) : initialize an iData array
• real b = real(s) : real value of iData object
• reshape c = reshape(a) : reshape the object Signal
• round b = round(s) : upper integer round of iData object
• set [s,...] = set(s, 'PropertyName', Propertyvalue, ...) : set iData properties
• single d = single(s) : convert iData into single floats
• size size(s) : get iData object size (number of elements)
• sort s = sort(a,dim) : Sort iData objects axes in ascending or descending order
• sqrt b = sqrt(s) : square root value of iData object
• std [half_width, center] = std(s, dim) : standard deviation of iData
• subsref b = subsref(a,s) : iData indexed references
• sum s = sum(a,dim) : computes the sum of iData objects elements
• unique s = unique(a,dim) : set unique iData objects axes with no repetitions
• zeros s = zeros(s,N,M,P,...) : initialize an iData array