stephenbeckr / sparsifiedkmeans Goto Github PK

Hello, when I try to run your demo, I have come across a problem. When I run the file 'example_loadFromDisk.m', the bug shows that Y = [Y Yj]; in file sampleAndMixFromLargeFile.m that the dim of Y and Yj is not equal. Could you help me with this problem? Thanks a lot.

Hey, dear stephen
When I run 'setup_kmeans', 'hadamard_pthreads.c' can't compile successfully, because “pthread.h”: No such file or directory.
Can you tell me if it is a real missing file or I can simply remove this line from code?
Best regards.

Hi!

I was using your sparsified kmeans and i think it is absolutely great. But i had encountered this error when using your code

Input: 5244949 x 35 in a mat file

Output: Splitting 5244949 x 35 matrix into Inf 5244949 x 0 chunks

The error code goes like this in matlab

`Warning: FOR loop index is too large. Truncating to 9223372036854775807.

In sampleAndMixFromLargeFile (line 96)
In kmeans_sparsified (line 301)
In example_loadFromDisk (line 55)
Error using sampleAndMixFromLargeFile (line 100)
Cannot index into 'X' because indices cannot be empty.

Error in kmeans_sparsified (line 301)
[X,timeLoad,timeMix,timeSample] = sampleAndMixFromLargeFile( DataFile, SparsityLevel, mix, p2,...

Error in example_loadFromDisk (line 55)
[indx,centers,sumd,dist_sketch] = kmeans_sparsified( [], k,'ColumnSamples',true,... '

I am using the loadfromDisk function but i've also tried using smaller values and it works like a charm.

To prevent small_p from being 0, maybe it's better to replace ln. 326 in kmeans_sparsified with:
small_p = max(1, round( SparsityLevel*p2 ));

(similarly ln. 323)

Hi~
I want to use a specific distance function (not provided in Matlab's original k-means) due to the structure of data I am dealing with. Since the distance function can simply be written as "distance = function(point_1, point_2)", I am wondering if customized distance function can be used in your wonderful implementation of K-means algorithm. Thank you~