Simitar Toolbox

A release of the Botvinick lab at the Princeton Neuroscience Institute, written by [email protected]. This fork is maintained by Chris Cox at the University of Wisconsin-Madison, [email protected].

Installation:

1. Add the toolbox directory to your MATLAB path ( addpath(<directory>) ). This will also work in Octave.

2. From MATLAB, go to the toolbox directory and try

	mex findNeighbours.c
	mex simitar.c
	mex fastscoring.c

If these work you will have much faster code for

• preparing metas with createMetaFromMask
• computing searchlight correlation matrices
• computing similarity structure score maps

That's it! If you want to learn more please follow the tutorial available on the Simitar web page.

Data preparation:

In order to use any functions with your own data you will need to create a data structure that maintains spatial information such as which voxels are neighbours of which. There is a page on this topic linked to from the tutorial page, or you can read README.datapreparation.txt, as it will explain how to create this data structure from a binary mask indicating which voxels are in the brain.

Functionality:

• Compute a searchlight similarity measure for the neighbourhood patch of each voxel function: computeSimilarityMap.m

• Scan for a particular similarity structure across all neighbourhood patches produced function: computeSimilarityStructureMap.m

• Compute similarity measure between two sets of examples (euclidean, correlation, cosine and others) function: computeExampleSimilarity.m

• Find the voxels that are in a given neighbourhood of each voxel (right now only supports cubic) function: computeNeighboursWithinRadius.m