[pymvpa] Cross-Modalities Searchlight
salim_alwasity at yahoo.com
Tue Jan 24 22:58:54 UTC 2017
DearsThe structure of my dataset has 2 Modalities BY 2 Actions design (2 Modalities "Motor Imagery (MI) and Action Observation (AO)") (2 Actions lift and knock).
I am trying to run a Searchlight to identify the common informative voxels (similar patterns) across different modalities, in another word, training the classifier on data of one modality (MI of Lift vs Knock) and tested on the other modality (AO of Lift vs Knock) and vice versa.Here is my code: clf=LinearCSVMC()
cv = CrossValidation(clf, HalfPartitioner(attr='modality'))sl = sphere_searchlight(cv, radius=3, postproc=mean_sample())res = sl(ds)
Where ds has a dimension of (32, 106720)(ds.sa.modality) return an array = ['MI' 'AO' 'MI' 'AO'.............................]which corresponding to (ds.sa.targets) = ['lift_nv' 'lift_v' 'knock_nv' 'knock_v'...........................]
However (res.samples) returns only 1s. Means nothing is common at all!!!Is there anything wrong with my code above?
Your comments are highly appreciated.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Pkg-ExpPsy-PyMVPA