The result folders are of the form:

SENS_<network_name>_<group_name>.txt

eg.: network_name=mbnetwork_falseids.txt_1_double 
and group_name=gr12intop40_double

and files in SENS_GROUP1 (files in SENS_GROUP2 are similar):

------------------------------------------
*.dat data files:
GR_<network_name>_<group_name>_0.15_Nr_Fm.dat
GR_<network_name>_<group_name>_0.15_Nr_Fp.dat

with Nr=group_size, e.g Nr=24 for gr12intop40_double

Data files with 5 columns for the quantities:
F_-(a,b) (Fm) and F_+(a,b) (Fp) being:
a, b, value of F_{+/-}(a,b), name of node a, name of node b
(4th and 5th columns are separated by a "#" symbol)
a,b=1,...,N with N=127 represent the matrix indicies

------------------------------------------
triangular graphique files: 
(red=positive maximum, green postiive intermediate, blue close to zero,
cyan negative intermediate, yellow strongest negative value)
GR_<network_name>_<group_name>_0.15_Nr_Fm_triag.png
GR_<network_name>_<group_name>_0.15_Nr_Fp_triag.png

full graphique file of initial GR matrix
GR_<network_name>_<group_name>_0.15_Nr.png 
(similar as alread present in corresponding GR_* folder)

vertical axis: a=1 (top) to a=N (bottom)
horizonal axis: b=1 (left) to b=N (right)

associated color bar file:  fig_bar.pdf
with numbers corresponding to F_{+/-}(a,b)/max_value 
(values with sign !)

------------------------------------------
*.mat files = initial data files used for creation of png files:
(format with N^2+2 lines being: 
N^2, N, N^2 data values in order (1,1), (1,2), ... )

GR_<network_name>_<group_name>_0.15_Nr_Fp_triag.mat
GR_<network_name>_<group_name>_0.15_Nr_Fm_triag.mat
GR_<network_name>_<group_name>_0.15_Nr_SD.mat
GR_<network_name>_<group_name>_0.15_Nr_SD2.mat
GR_<network_name>_<group_name>_0.15_Nr.mat
(These files are not very important for analysis, better use above 
"*.dat" files with 5 columns).
------------------------------------------
data file of used node names in PageRank order (of simple network):

<group_name>.txt_names
------------------------------------------

Shown quantities:

GR(a,b) = reduced Google matrix of (simple) network and Group1 or Group2)
(see Eq. (8) of Plos paper of 2018 (my253.pdf)).

Sensitivity given by Eq. (14) of Plos paper:
D_ab(x)=\lim_{\eps\to 0} 1/P(x) dP(x)/d\eps
where GR is perturbed by GR(a,b)-> GR(a,b)*(1+\eps) for (a,b) element 
and no modification of other elements

Using D_ab(x) one defines the symmetric version (see Eq. (15) of Plos paper):

DS_ab(x)=D_ab(x)+D_ba(x)

and using this one defines:

F_+(a,b)=DS_ab(a)+DS_ab(b)
F_-(a,b)=DS_ab(a)-DS_ab(b)

(see Eq. (16) of Plos paper for F_-(a,b) being F(a,b) in (16); the 
quantity F_+(a,b) is a new quantity)

The latter two are shown in files with "Fp" or "Fm" in file name.
---------------------------------------

The file with "SD.mat" in the name show:

SD(a,b) defined as: SD(a,b)=D_ab(a) with D_ab(x) as above.


The file with "SD2.mat" in the name show:

SD2(a,b) defined as: SD2(a,b)=SD(a,b)/P(b)

Remark: generically SD(a,b) is strongly proportional to P(b) (except 
of a few number of pics etc.)

If one use a toy-model for GR with columns proportional to PageRank P 
(i.e. with "only" Gpr-contribution and vanishing Grr, Gqr) 
one can analytically show that:

SD(a,b) = (1-P(a))*P(b) \approx P(b) since typically P(a)<<1

and SD2(a,b) = 1 - P(a) \approx 1 since typically P(a)<<1 


