This file contains information to introduce you to the Clustering Library. Please note that this library is not formally released, and is therefore not supported by National Instruments. This library is a set of VIs and demos designed as starting points for clustering analysis in LabVIEW.
Overview
The goal of clustering is to partition data samples into different classes according to the samples' mutual similarity. First, a distance metric is defined to measure similarity. Next, by minimizing the overall cost, you can determine the centroids and partition scheme. This process is often iterative and the cluster centroids will converge after several iterations. Possible applications of clustering include pattern recognition, data compression, image segmentation, and so on.
Download the attchement below and let us know what you think. We'd love to get your feedback!
It is some Joke ??
Clustering is a very useful tool, especially in regards to Condition Monitoring and Process Monitoring. Many physical measurements are made, and then features / metrics / calculations from these measurements. The results of these calculations on the measurements then form a "state" vector of the machine, machine component, or process. This vector can then be clustered into know operational and failure modes of the machine or process.
ha! No, this has nothing to do with creating a cluster in LabVIEW if that is what you were thinking.
Hi,
I am looking for clustering algorithms for image processing pattern recognition.
What clustering algorithms do you have in this library?
Thanks - Amit
Hi,
What algorithm would you like to use? There are many algorithms in the library including K-means, Fuzzy K-means..
ZJ Gu
Hi,
I have K-mean that I have wrote a few years ago. I will try the library and see what I get with the other algorithms.
Do you have also classification library? SVM, fisher discriminant, etc.'...
Thanks - Amit,
Hi Amit,
In the current library , SVM is not included. You could install it and have a try.
Thanks!
Fantastic algorithm. Thanks very much. I am processing data using K-means. Each channel has about 60000 data points and there are 18 channels to be grouped into 5 clusters... With this large data set, the k-means algorithm froze in the dot product subvi and was generally slow. I replaced that dot product subvi with "multiplication" and "add array elements". The clustering speed was much faster. Try it, see if it speed things up for you also. Other than that, it working good for me. Here are the changes I made.