Log in Page Discussion History Go to the site toolbox

Automatic Clustering

From BluWiki

KlustaKwik

Developed by Ken Harris. See the Klustakwik homepage for details.

The following is an example on how to run KK in a machine called bard02.cshl.edu. These scripts for preparing the data, running the clustering algorithm, and generating reports, are still under development (as of June 2007). Please contact Santiago in case of errors or bugs.

Let's assume you have NSpike data for two tetrodes:

tmpdata007
    |__ 01-000
    |    |__ 01-000.tt
    |
    |__ 02-000
         |__ 02-000.tt
  • First, you need access to bard02.cshl.edu (IP: 143.48.30.47), as user bard. Ask Santiago for a password.
  • If your data resides on a Windows machine, you will need two additional tools:
    • PuTTY to be able to run commands in bard02.cshl.edu remotely.
    • WinSCP to be able to copy files to bard02.cshl.edu.
  • If your data resides in a Linux/Unix machine (for example nspikemaster), see Clustering instructions for those new to command-line LINUX.
  • Copy your NSpike data to the directory /home/bard/data/ of bard02.cshl.edu.
  • Log into bard02.cshl.edu as user bard.
  • Run Matlab with no java virtual machine: matlab -nojvm -nodisplay
    • If it complains, make sure the license manager is running. Run: /usr/local/matlabR2006b/etc/lmstart
  • Within matlab:
createfetfiles('/home/bard/data/santiago/tmpdata007/',[1,2]);
runclustering('/home/bard/data/santiago/tmpdata007_kk',[1,2]);

The first line creates the directory tmpdata007_kk and saves waveform features in a format that KlustaKwik can read. The second line runs KK on tetrodes 1 and 2 in parallel. These processes run in the background, so you can keep working in matlab, but you shouldn't run the next step until they finish.

Optionally, you can exclude the signals from particular leads from being used by the clustering algorithm (which you might want to do if you know you have a bad lead). See the help in runclustering for details.

The last step is to generate reports (in PostScript) about the resulting clusters:

printclusterreport('/home/bard/data/santiago/tmpdata007',[1,2],'/home/bard/data/santiago/tmpdata007_report')

Watershed

Developed by Alex Koulakov: Watershed-based spike sorting. Link title

Site Toolbox:

Personal tools
GNU Free Documentation License 1.2
This page was last modified on 24 January 2008, at 22:59.
Disclaimers - About BluWiki