flowstab.scripts.run_clusterings
Clusters the covariance integrals computed with run_cov_integrals.py (step 2).
The clusterings are computed with the Louvain algorithm for each file containing the result of step 2 in {savedir}.
The Louvain algorithm is repeated num_repeat times and the best partition (maximizing the flow stability) is selected.
The ensemble of partitions is used to compute the variation of information.
The computations for the different files is spread of nproc_files in parallel. It is also possible to parallelize the
repeated clusterings using nproc_clustering. This requires nproc_files*nproc_clustering cores.
A number of metadata (e.g. all the parameters used to run the code) is saved together with the results in a dictionary.
The results of the clustering are saved with the following keys:
- ‘clust_counter’ : All the different partitions found at each run of the Louvain algorithm togheter with the number of times they appeared.
- ‘stabilites’ : All the values of the flow stability for each run.
- ‘seeds’ : Random seeds for each run.
- ‘nvarinf’ : Average normalized variation of information computed over the ensemble of partitions.
- ‘avg_stab’: Average flow stability computed over the ensemble of partitions.
- ‘avg_nclust’ : Average number of clusters computed over the ensemble of partitions.
- ‘best_cluster’ : Best partition (the one with the max flow stability from the ensemble).
- ‘best_stab’ : Flow stability value of the best partition.
The results are saved in files named f’{savedir}/clusters_{netname}_tau_w{tau_w:.3e}_PT_{int_start:06d}_to_{int_stop:06d}.pickle’
Module Contents
-
flowstab.scripts.run_clusterings.compute_nvi_sample(params)[source]
-
flowstab.scripts.run_clusterings.compute_static_clustering_fct(params)[source]
-
flowstab.scripts.run_clusterings.load_autocov_int(file, init_p1=False, direction=None)[source]
-
flowstab.scripts.run_clusterings.main()[source]
-
flowstab.scripts.run_clusterings.n_random_seeds(n)[source]
-
flowstab.scripts.run_clusterings.worker(file_args)[source]
-
flowstab.scripts.run_clusterings.ap[source]
-
flowstab.scripts.run_clusterings.clust_verbose[source]
-
flowstab.scripts.run_clusterings.compute_clustering[source]
-
flowstab.scripts.run_clusterings.compute_nvarinf[source]
-
flowstab.scripts.run_clusterings.compute_static_clustering[source]
-
flowstab.scripts.run_clusterings.datadir[source]
-
flowstab.scripts.run_clusterings.direction[source]
-
flowstab.scripts.run_clusterings.files[source]
-
flowstab.scripts.run_clusterings.files[source]
-
flowstab.scripts.run_clusterings.inargs[source]
-
flowstab.scripts.run_clusterings.init_p1[source]
-
flowstab.scripts.run_clusterings.integral_rtol[source]
-
flowstab.scripts.run_clusterings.max_num_varinf_samples[source]
-
flowstab.scripts.run_clusterings.n_meta_iter_max[source]
-
flowstab.scripts.run_clusterings.n_sub_iter_max[source]
-
flowstab.scripts.run_clusterings.net_name[source]
-
flowstab.scripts.run_clusterings.normalize_cov_mat[source]
-
flowstab.scripts.run_clusterings.nproc_clustering[source]
-
flowstab.scripts.run_clusterings.nproc_files[source]
-
flowstab.scripts.run_clusterings.num_norm_iter[source]
-
flowstab.scripts.run_clusterings.num_repeat[source]
-
flowstab.scripts.run_clusterings.num_varinf_samples[source]
-
flowstab.scripts.run_clusterings.num_varinf_samples[source]
-
flowstab.scripts.run_clusterings.only_largest_comp[source]
-
flowstab.scripts.run_clusterings.optional[source]
-
flowstab.scripts.run_clusterings.print_num_loops[source]
-
flowstab.scripts.run_clusterings.required[source]
-
flowstab.scripts.run_clusterings.save_clustering_before_nvarinf[source]
-
flowstab.scripts.run_clusterings.savedir[source]
-
flowstab.scripts.run_clusterings.separate_comps[source]
-
flowstab.scripts.run_clusterings.var_dict[source]
-
flowstab.scripts.run_clusterings.verbose[source]