top of page
Proposed Algorithm: FastPole
Visit our algorithm on Github:
Whole Logic Structure
C: candidates of multipoles, c∈C
Σ：correlation matrix of fMRI data
σ*: threshold value of linear dependence
δ*: threshold value of linear gain
M: array built to store multipoles
S_max: largest subsets of c
Check Maximal Clqiues
Step 1:
Calculate dependence of candidates, σ_c
Step 2:
-
Calculate dependence of each subset(length=|S|-1)
-
Store the dependences in a tuple array, S_max
-
Sort S_max by dependence values in a descending order
Step 3:
-
Use σ_c and S_max to compute the gain of the candidate
-
•Push multipoles into M
Check Sub Clique
Step 1:
-
Compute the upperbound of multipole size, l_max
-
If l_max=|c|-1
Push subsets (dependence>=sigma) in S_max to S'_max
Else
S'_max={Combinations(c,l_max),dependences}
Step 2:
-
Compute the dependence of subsets (length=|S'_max[1]|-1) and store them in a tuple array, S_min
-
Push subsets (dependence>=sigma) in S_min to S'_min
Step 3:
-
Use S_max and S'_max to compute the gain of each element in S_max
-
Once find multipoles, push multipoles, end; Otherwise, Sup=sub, return to Step 2
bottom of page