Spike classifiers


Jump to: navigation, search

We are implementing a variety of classifiers that run on spiking data. These include classifiers that

  • convert spikes into rates (continuous values) and submit the data to SVM classifiers (libSVM),
  • take spikes directly and learn to classify
  • liquid state machines

Return to Synapse page


[edit] Installation

The classify library routines are used to make decisions about the spiking outputs.

There are three possible implementations of an SVM:

  1. svmtrain/svmclassify from the the bioInformatics toolbox
  2. a matlab implementation of svm
  3. the svmlib interface for Matlab

Mostly, we have been using libsvm. You can download libsvm.

[edit] Linux

There are mex files for R2008b in the svn repository. We think these also run for R2008a. If you need to reinstall libsvm

  • Find your matlab directory for the version you want to use (it has a bin/ sub-directory)
  • Modify the Makefile file accordingly (someone provide a hint here?)
  • Then type make

[edit] Windows

  • See the README file (provide a hint here?)

In both cases, we suggest renaming mex files svmtrain.* to libsvm_svmtrain.* This avoids a possible name conflict with the Bioinformatics toolbox.

[edit] Implementation of a spiking classifier

Personal tools