Bayesian component clustering

[arnodelorme]

Get a probability for each cluster.

• loop bootstrap trials - loop scan measures - compute cluster using different measures (ERP, spectrum, dipoles, etc...) - in cluster matrix of components x components, add 1 if 2 components are in the same cluster
• Look at the matrix components x components. Then you get the likelihood for 2 components to be in the same cluster.
• Scan each component, set a threshold (75% for example) for belonging to the same cluster with others.
• Now you have your cluster

@CPernet idea

[CPernet]

there is some code somewhere from @nucleuscub