Figure 4.

The extension of the generative model in Figure 1 to the fusion of EEG and MEG data, i. e., j = 1–2 modalities (Y₁ and Y₂). This model was the one used to produce the results in Figure 5; i.e., with one minimum norm source prior (${\mathbf{\text{Q}}}_1^{(2)}={\text{I}}_p$, where I_p is a p × p identity matrix for the p sources) and one white-noise sensor component for each modality (${\mathbf{\text{Q}}}_j^{(1)}={\text{I}}_{n_j}$, where n_j is the number of sensors for the j-th sensor-type).