Skip to main content
. 2017 Aug 18;6:1512. [Version 1] doi: 10.12688/f1000research.12353.1

Figure 8. Experimental results for decentralized independent vector analysis (dIVA) ( Wojtalewicz et al., 2017).

Figure 8.

The experiment is based on synthetic data using a generalized autoregressive conditional heteroscedatic model and the SimTB functional MRI Simulation Toolbox ( Erhardt et al., 2012). The top figure shows how the processing time, number of iterations, and intersymbol interference (ISI) change as the global number of subjects increases. The processing time increases with the number of subjects per site ( A). Additionally, feature quality increases, indicated as decreasing ISI ( C). The bottom figure shows the processing time ratio between dIVA and IVA decreases as the global number of subjects increases. When the global number of subjects reaches 512, dIVA requires only one quarter of the processing time of IVA.