search

BIAPT / awareness-perturbation-complexity-index

Development of an index for assessing the level of consciousness of healthy and disorder of consciousness individuals.
1 stars 0 forks source link

Calculate the cosine similarity between each of the three states #26

Closed yacineMahdid closed 4 years ago

yacineMahdid commented 4 years ago

Like we did in the dpli-dri we should generate the cosine similarity between each of the three states in order to have a numerical values representing the difference. We should do that for each of the participants.

Need to generate the hubs -> then cosine similarity at different thresholds:

yacineMahdid commented 4 years ago

There are thresholds here since we are working with the weights matrix.

yacineMahdid commented 4 years ago

Also, need to filter the matrices beforehand to have the same size otherwise we cannot calculate the cosine similarity

yacineMahdid commented 4 years ago

Some refactoring will need to be done for get_subset. We should be able to iterate over an arbitrary amount of locations.

yacineMahdid commented 4 years ago

Will need to either remove participant WSA17 from the analysis or refactor the code to automatically take his presence into consideration. However, I don't see any value added by having him in the analysis since we do not have the recovery for this participant.

yacineMahdid commented 4 years ago

I have to be careful with labels and full blown location, this is creating some problem in the hub location as the channels location are an important source of information. For the sake of time I will not change get_subset, instead I will regenerate the channels information from the labels. I should make a better get_subset afterward where we have correct parameters.

yacineMahdid commented 4 years ago

Need also to fix up the documentation for the helper function and to normalize the whole thing. Maybe by putting them into NeuroAlgo utils folder.