Algorithm 1. Pseudo Code for Iterative Matrix Decomposition.

Inputs: MEG sensors space matrix $X\in {\mathbf{\Re }}^{N\times M}$

Process:

1. For i = 1,…,R

Randomly initialize t

$\Vert \mathit{t}\Vert \to 1$, $\widehat{\mathit{t}}=2\mathit{t}$

While $\Vert \widehat{\mathit{t}}-\mathit{t}\Vert >\mathrm{eps}$

$\widehat{\mathit{t}}=\mathit{t}$

$\mathit{p}=X^{\mathit{T}}\mathit{t}$

$\Vert \mathit{p}\Vert \to 1$

$\mathit{t}=X\mathit{p}$

$\Vert \mathit{t}\Vert \to 1$

End

Extract the score and loading vector t, p

Deflate X matrix: $X=X-\mathit{t}{\mathit{p}}^{\mathit{T}}$

End

2. $T=[{\mathit{t}}_1,{\mathit{t}}_2,...,{\mathit{t}}_{\mathit{R}}],P=[{\mathit{p}}_1,{\mathit{p}}_2,...,{\mathit{p}}_{\mathit{R}}]$

3. Reconstruct the sensors matrix $\widehat{X}=T{P}^{\mathit{T}}$

Outputs: Score matrix T, loading matrix P, and reconstruction matrix $\widehat{X}$.