dataset - How to generate multidimensional data with specific clustering properties? -

-

in section 5.a of research paper researcher used following synthetic datasets:

  1. gauss consisted of 6 gaussian clusters identity covariance, each 500 points in 5 dimensions. means randomly assigned value 0 10 in each dimension. cluster means required @ least 4 euclidean distance apart, , points required within 2 euclidean distance of cluster mean.

  2. paired consisted of 3 pairs of gaussian clusters identity covariance, each 500 points in 5 dimensions. each pair of gaussians placed around mean randomly assigned value in each dimension 0 20 such euclidean distance between paired gaussian clusters between 4 , eight, , euclidean distance between non-paired gaussians @ least 12. additionally, points required within 2 euclidean distance of cluster mean.

  3. elong consisted of 5 gaussian clusters identity covariance, each 300 points in 5 dimensions. means randomly assigned value 0 50 in each dimension. create elongated clusters in different dimensions, multiplied values of single, distinct dimension each cluster 15. cluster means required @ least 5 euclidean distance apart.

  4. uniform consisted of 8 clusters, each 300 points in 3 dimensions. each cluster had points uniformly distributed in 3x3x3 box around randomly assigned center in 10x10x10 cube. cluster centers required 5 euclidean distance apart.
  5. rings consisted of 2 ring clusters centered around (0,0), larger outer ring radius 2 , smaller inner ring of radius 1. 400 points evenly spaced degrees on inner ring.

http://postimg.org/image/jo4rjztjz/

i don't have these datasets. tried contact researcher of no use.

how create these datasets? there kind of tool create them?

original paper can found here

documentation , examples on elki data set generator can found here: http://elki.dbs.ifi.lmu.de/wiki/datasetgenerator

the generator in elki cannot produce ring-shaped clusters (only spherical), , not support clipping points @ maximum distance. generates independent samples each dimension independently. supported operation uses more 1 dimension @ time rotation operation. generating ring-shaped clusters, or clipping clusters based on distance mean means form of dependence of values not supported.

you need either contact authors of publication, or write program generate such data yourself. it's not hard; may not worth effort generate such synthetic data - it's not realistic scenario in opinion.


Comments

Post a Comment

Popular posts from this blog

c++ - Delete matches in OpenCV (Keypoints and descriptors) -

-
i want check scene image against 2 train images. that, detect features , compute descriptors of both training images. before detecting, computing , matching scene image, delete matches of train1 , train2. because these matches won't facilitate matching of scene image train1 , train2. so, match train1 train2 , vector of matches trainidx , queryidx. how can delete these matches in keypoints-vector , descriptor matrix of train1 , train2? best regards, dwi i have done below: std::vector<cv::keypoint> keypoints[2]; cv::mat descriptor[2]; std::vector< cv::dmatch > matches; /* write code generate keypoints, descriptors , matches here... keypoint[0] -> train image 1 keypoints keypoint[1] -> train image 2 keypoints descriptor[0] -> train image 1 descriptors descriptor[1] -> train image 2 descriptors matches -> matched between train image 1 , 2 */ // logic keep unmatched keypoints , corresponding descriptor
Read more

java - Could not locate OpenAL library -

-
i writing game using slick2d library, uses lwjgl. lwjgl version 2.9.3. using netbeans 8. i tried adding sounds game, example music track on main menu. problem is, exception when run game, despite game still compiling , running except sound of course: sat mar 21 01:23:57 edt 2015 error:could not locate openal library. org.lwjgl.lwjglexception: not locate openal library. @ org.lwjgl.openal.al.create(al.java:156) @ org.lwjgl.openal.al.create(al.java:156) @ org.lwjgl.openal.al.create(al.java:102) @ org.lwjgl.openal.al.create(al.java:206) @ org.newdawn.slick.openal.soundstore$1.run(soundstore.java:295) @ java.security.accesscontroller.doprivileged(native method) @ org.newdawn.slick.openal.soundstore.init(soundstore.java:292) @ org.newdawn.slick.music.<init>(music.java:156) @ org.newdawn.slick.music.<init>(music.java:75) @ view.mainmenu.init(mainmenu.java:58) i didn't include rest of error states error happens in game code. the
Read more

sorting - opencl Bitonic sort with 64 bits keys -

-
i used nvidia sdk bitonic sort , works great me. 32 bits (uint) need ulong keys. have typically 2^14 keys @ time , power of 2. searched on not find kernel designed ulong. i tried modified nvidia sdk bitonic sort using ulong keys not work. kernel not crash after de call clenqueuendrangekernel got error : cl_invalid_command_queue can tell me how modify bitonic sort or radixsort, or can sort ulong keys? i running nvidia quadro 4000 opencl 1.1 cuda 6.5.20 full_profile used original nvidia sdk bitonicsort.cl i used"ulong" keys , value input , ouput instead of "uint" thanks helping i made work-around suit need. used 30 bits of key , 18 high bits of value sorting. changed comparators (2 lines): '#define mask 0xffffc000 // suit need from this: if( (*keya > *keyb) == dir ) this: if(((*keya > *keyb)||((*keya == *keyb)&&((*vala & mask)> (*valb & mask)))) == dir ) it works fine. but, still, it's work-a
Read more