Cluster configurations of the Deffuant model on network ensembles

## Data

For each measured combination of the confidence and system size, there is one gzipped
file. For different ensembles, we collected data in different ranges and quality.
The paramters are:

* Number of samples `m` per parameter combination
* Range `r` of confidences epsilon
* Distances `d` between values of epsilon (basically the resolution of the data)
* Largest size `N_max`

The single files follow a naming scheme of `n{N}_e{epsilon}.cluster.dat.gz`, where
`{N}` signals the system size of the simulation and `{epsilon}` is the confidence
value of the simulation (without a decimal point, i.e., `0050` corresponds to `epsilon = 0.050`).
The sizes `N` are usually powers of two (or for the lattices, perfect squares close to powers of two).

We present the data for each ensemble in one folder (after unpacking the tar archive).
Note that some parameter values are missing, if they did not converge in reasonable time.

* Fully connected `full`
    * `m = 1000`, `r = [0.0, 0.6]`, `d = 0.001`, `N_max = 1048576`
* Barabasi Albert with a mean degree of 10 `BA10`
    * `m = 1000`, `r = [0.0, 0.6]`, `d = 0.002`, `N_max = 16384`
* Square lattice with third nearest neighbors `lat3`
    * `m = 1000`, `r = [0.0, 0.6]`, `d = 0.002`, `N_max = 4096`
* connected Erdos Renyi with mean degree of 10 `ER10`
    * `m = 1000`, `r = [0.0, 0.3]`, `d = 0.002`, `N_max = 16384`

## Data format

Each final state is encoded as three lines:

* The convergence time is a single integer with a line prefix '# sweeps: '
* The positions of all clusters in opinion space with a line prefix '# ' (unsorted)
* The number of agents in each of the clusters without a line prefix

## Python example for reading the format

An example script, which visualizes the S vs eps graph for the largest size of the fully connected
case, with a function to read this format is given in `example.py`.