KlustaWin is a Windows implementation of version 1.6 of the program KlustaKwik , which is written by Dr. Ken Harris (e-mail: harris@axon.rutgers.edu).
The following details of KlustaKwik are taken from Dr Harris's notes.
KlustaKwik is a program for unsupervised classification of multidimensional continuous data. It arose from a specific need - automatic sorting of neuronal action potential waveforms (see KD Harris et al, Journal of Neurophysiology 84:401-414,2000). We needed a program that would
1) Fit a mixture of Gaussians with unconstrained covariance matrices
2) Automatically choose the number of mixture components
3) Be robust against noise
4) Run fast on large data sets (up to 100000 points, 48 dimensions)
Speed in particular was essential. KlustaKwik is based on the CEM algorithm of Celeux and Govaert (which is faster than the standard EM algorithm), and also uses several tricks to improve execution speed while maintaining good performance.
What this means is that KlustaWin can be used to analyse populations with n-dimensional parameters that are made up of mixtures of sub-populations, so that these sub-populations are classified into separate clusters. The parameters each have a normal distribution with their own mean and standard deviation, but the resulting clusters do not have to have a spherical distribution (unlike, e.g., K-means clustering). A typical use is in analysing extracellular recordings of neural action potentials. Such recordings often contain waveforms derived from several different axons, but each axon produces a waveform with approximately the same shape. Principal component analysis is often used to produce a numerical description of each waveform, usually consisting of 3 separate numbers for each waveform – i.e. a 3-dimensional array of numbers. The numerical values from particular axons tend to cluster around each other in a 3-dimensional “cloud”. KlustaWin is used to classify these numerical values into their appropriate clusters.
The source code for KlustaKwik and KlustaWin is available under the GNU General Public Licence (“www.gnu.org”).
KlustaWin builds on the KlustaKwik code by implementing a graphical user interface with 3-D visualization of data, but does not change the clustering algorithm, and makes only superficial changes to the source code of that algorithm. Therefore all queries regarding the algorithm or its implementation should be directed to the author of KlustaKwik.
Here are the really simple instructions:
More details are given in the sections below.