FIGURE 2.

Snowball ICA algorithm. In order to avoid the necessity of specifying model order in conventional ICA, Snowball ICA estimates components one by one (or a few at a time) until no stable components can be further estimated from the data. After seed creation for a new component(s) to be estimated via information collection, the estimated snowball component(s) are removed from the multi-subject data and seed creation is repeated to determine the next seed pattern(s). The Snowball ICA will stop when there are no more stable components estimated during the seed creation step.