5.1.2 Data set (B)

Data set (B) differs from data set (A) only by having some processes which are subject specific, i.e. they have zero amplitude modulation in some of the subjects' data. Figure 3 shows the results for the 'common' source process (top) and one of the secondary source process (bottom) which only exists in subject 2. The model order was estimated to be $R=13$. Both techniques still estimate the source processes, with the PARAFAC estimate of the common source process showing some cross-talk with spatial map 2. The PARAFAC estimation in this case involved a 23-fold increase in number of floating point operations.

Figure 3: Tensor-PICA and PARAFAC decomposition results for data set (B) with $R=13$: the first process is contained in all three subjects while the second process only appears in subject 2.

(i) tensor-PICA (ii) PARAFAC