def opt_fn(work_par):
work_par = work_par[0]
if method in ('ilfrs', 'ilfr'):
partition = community_ext.best_partition(G,
model=method,
pars={'mu': work_par})
loglike = community_ext.model_log_likelihood(G,
partition,
model=method,
pars={'mu': work_par})
else:
partition = community_ext.best_partition(G,
model=method,
pars={'gamma': work_par})
loglike = community_ext.model_log_likelihood(
G, partition, model=method, pars={'gamma': work_par})
return -loglike
def opt_par(G, method, par):
def opt_fn(work_par):
work_par = work_par[0]
if method in ('ilfrs', 'ilfr'):
partition = community_ext.best_partition(G,
model=method,
pars={'mu': work_par})
loglike = community_ext.model_log_likelihood(G,
partition,
model=method,
pars={'mu': work_par})
else:
partition = community_ext.best_partition(G,
model=method,
pars={'gamma': work_par})
loglike = community_ext.model_log_likelihood(
G, partition, model=method, pars={'gamma': work_par})
return -loglike
# find the optimal parameter value
best_par = float(
scipy.optimize.fmin_powell(opt_fn, par, full_output=False, disp=False))
# calculate the partition with the optimal parameter value
if method in ('ilfrs', 'ilfr'):
partition = community_ext.best_partition(G,
model=method,
pars={'mu': best_par})
loglike = community_ext.model_log_likelihood(G,
partition,
model=method,
pars={'mu': best_par})
else:
partition = community_ext.best_partition(G,
model=method,
pars={'gamma': best_par})
loglike = community_ext.model_log_likelihood(G,
partition,
model=method,
pars={'gamma': best_par})
return partition, loglike, best_par
model=method,
pars={'mu': work_par})
else:
partition = community_ext.best_partition(G,
model=method,
pars={'gamma': work_par})
# calculate optimal parameter value for the current partition
if method in ('ilfrs', ):
work_par = community_ext.estimate_mu(G, partition)
else:
work_par = community_ext.estimate_gamma(G, partition, model=method)
loglike = community_ext.model_log_likelihood(G,
partition,
model=method,
pars={
'gamma': work_par,
'mu': work_par
})
print('current par', work_par, 'loglike', loglike)
# calculate and print the scores of resulting partition
part_scores = community_ext.compare_partitions(groundtruth_partition,
partition)
loglike = community_ext.model_log_likelihood(G,
partition,
model=method,
pars={
'gamma': work_par,
'mu': work_par
})