Algorithm 1.

The V B pseudocode

Input: Multichannel EEG data Xk (k = 1, 2) that are recorded from two experimental conditions

Output: The variational distributions q*(Zk), q*(A), q*(α), $q^{*}({[{\mathbf{\Lambda }}_k^{(i)}]}^{-1}),q^{*}({\mathbf{\Psi }}_k^{-1})$

Initialization: Use the settings in (14), and set iter = 0, ℒ(0) = −inf, $u^{(m)}={\upsilon }^{(m)}=e_k^{(m)}=f_k^{(m)}=g_k^{(c)}=h_k^{(c)}={10}^{-8}$ (c = 1, …, C; k = 1, 2; m = 1, …, M)

Method:

repeat

iter = iter + 1

Update the parameters in the variational distributions using (7) – (12)

if (iter mod 100) = 0 then

Compute ${\{e_k^{(m)},f_k^{(m)}\}}_{k=1,2;m=1,\dots ,M}$ by solving (13)

end if

Compute the variational lower bound ℒ(iter) using (19)

until ℒ(iter) − ℒ (iter − 1) ≤ a pre-defined threshold (e.g., 10−8)