Format of files for grpsan.f90:

Parameter file grpsan.par

13 lines define 13 variables.

N : number of data sets or rank of the correlation matrix.
filxxx : three letters prefix for names of input/output files.
k0 : number of reclustering step. 0 means first clustering.
clnm : this is 1 if the file with the name of the series exists. It has to be called xxx.names.
be_in : Initial beta (inverse temparature).
be_fin : Final beta (inverse temperature).
t_wri : number of sweeps between consecutive writings on file xxx.ent.gsK.
t_temp : number of sweeps between changes in temperature.
nr_f : number of annealing cycles. Each cycle lowers and increases the temperature. At the end there is a final low temperature run.
cf : factor for changes in beta (inverse temperature).
idum : NEGATIVE seed for random number (we use a routine similar to ran2 of numerical recipies).
flag : initial condition: 0=one set per cluster; 1=condition read from filgs=xxx.cnf.gs(k0-1); 2=random initial condition.
rns : 1 if random sweep of lattice (else sequential).

On input

xxx.cij: is the covariance matrix in upper triangular form. It is written as a single column of numbers, where rows of the matrix are appended. Only the elements on the right of the diagonal are reported (the diagonal is excluded, being made of 1s).

xxx.names: Two columns. The first has the number of the series and is not used in grpsan.f90. The second has the names of the series. Names are up to 18 characters long.

On output

xxx.ent.gsK: 8 columns. Status of the system during the annealing.

column 1 : time, counts the number of steps.
column 2 : inverse temperature (beta).
column 3 : energy (minus log of likelihood), which is being minimized.
column 4 : number of spin updates, per spin, per sweep, since last writing
column 5 : number of clusters
column 6 : average number of set per cluster
column 7 : number of merging, per sweep, since the last writing
colunm 8 : number of splitting, per sweep, since the last writing

xxx.cnf.gsK: spin configuration of the ground state. Best cluster structure. 1 column with the labels of the spins. It is written so that the first line is the number of the cluster to which the first dataset belongs, and so on...

xxx.cls.gsK: 3 columns. Ground state cluster structure. Clusters are separated by empty rows. Cluster are sorted according to their energy per spin e/n. Inside the cluster the sets are ordered according to their extraction energy.

column 1 : number of the data set.
column 2 : name of the data set (optional).
column 3 : energy required to extract the spin from the cluster.

xxx.clt.gsK: 3 columns. Same as xxx.cls.gsK but with informations on the clusters on top of each cluster. It is supposed to be human readable.

xxx.now.gsK : 3 columns.The ground state energy at this time. This file is not initialized, so that new runs append their data to it.

column 1 : Ground state energy (per spin).
column 2 : inverse temperature; 0 means initial condition.
column 3 : time; 0 means initial condition.