def t_selection_pipeline_directed_wiki(G,ts,num_partitions,fraction_t_to_keep=0.25):
mis = []
coups = []
d_gws = []
rt = []
for t in ts:
start = time.time()
cost = sgw.directed_heat_kernel(G,t)
mutual_info, d_gw, coup = process_sgwl_wiki(cost,database,num_nodes,num_partitions)
mis.append(mutual_info)
coups.append(coup)
d_gws.append(d_gw)
end = time.time()
rt.append(end-start)
print('Couplings Computed')
coverages = []
for j in range(len(ts)):
coup = coups[j]
partition = get_partition(coup)
coverages.append(coverage(G,partition))
num_to_keep = int(np.round(fraction_t_to_keep*len(ts)))
good_t_max = ts[np.argsort(coverages)][-num_to_keep:]
good_t_grad = ts[np.argsort(np.abs(np.gradient(coverages)))][:num_to_keep]
return mis, coups, d_gws, good_t_max, good_t_grad, rt
# print('--- Noisy data | Symmetrized | SpecGWL | Best mutual information score: {:3.3f} | @t = {:3.3f} | average runtime per iteration = {:3.3f}'.format(max(mis), ts[np.argmax(mis)], np.mean(rt)))
scores['specgwl-symmetrized-noisy'] = max(mis)
runtimes['specgwl-symmetrized-noisy'] = sum(rt)
# avetimes['specgwl-symmetrized-noisy'] = np.mean(rt)
###########################################################
###########################################################
# Proposed method: SpecGWL (asymmetric)
###########################################################
# Raw
mis = []
rt = []
ts = [10] #np.logspace(-3,1,10)
for t in ts:
start = time.time()
cost = sgw.directed_heat_kernel(dG,t)
mutual_info = process_sgwl_wiki(cost,database,num_nodes,num_partitions);
mis.append(mutual_info)
end = time.time()
rt.append(end-start)
# print('--- Raw data | Asymmetric | SpecGWL | Best mutual information score: {:3.3f} | @t = {:3.3f} | average runtime per iteration = {:3.3f}'.format(max(mis), ts[np.argmax(mis)], np.mean(rt)))
scores['specgwl-asymmetric-raw'] = max(mis)
runtimes['specgwl-asymmetric-raw'] = sum(rt)
# avetimes['specgwl-asymmetric-raw'] = np.mean(rt)
# Noisy
mis = []
rt = []
ts = [10] #np.logspace(-3,1,10)
for t in ts:
##############################################
# Training specGW
##############################################
G = nx.gaussian_random_partition_graph(n=num_nodes, s=clique_size, v=8,
p_in=p_in, p_out=ps_out[pn], directed=True)
p_s, cost_s, idx2node = DataIO.extract_graph_info(G)
p_s = (p_s + 1) ** 0.01
p_s /= np.sum(p_s)
start = time.time()
t = 10
cost = sgw.directed_heat_kernel(G,t)
modularities = []
for j in num_clusts:
p_t = GwGt.estimate_target_distribution({0: p_s}, dim_t=j)
sub_costs, sub_probs, sub_idx2nodes, coup, d_gw = graph_partition_gd2(cost,
p_s,
p_t,
idx2node,
ot_dict)
partition = get_partition(coup)
modularities.append(modularity(G,partition))
est_num_clust = num_clusts[np.argmax(modularities)]
# print('--- Noisy data | Symmetrized | SpecGWL | Best mutual information score: {:3.3f} | @t = {:3.3f} | average runtime per iteration = {:3.3f}'.format(max(mis), ts[np.argmax(mis)], np.mean(rt)))
scores['specgwl-symmetrized-noisy'] = max(mis)
runtimes['specgwl-symmetrized-noisy'] = sum(rt)
# avetimes['specgwl-symmetrized-noisy'] = np.mean(rt)
###########################################################
###########################################################
# Proposed method: SpecGWL (asymmetric)
###########################################################
# Raw
mis = []
rt = []
ts = [4] #np.logspace(-1,2,100)
for t in ts:
start = time.time()
cost = sgw.directed_heat_kernel(dG, t).astype(np.float64)
mutual_info = process_sgwl_wiki(cost,
database,
num_nodes,
num_partitions,
beta=2e-5)
mis.append(mutual_info)
end = time.time()
rt.append(end - start)
# print('--- Raw data | Asymmetric | SpecGWL | Best mutual information score: {:3.3f} | @t = {:3.3f} | average runtime per iteration = {:3.3f}'.format(max(mis), ts[np.argmax(mis)], np.mean(rt)))
scores['specgwl-asymmetric-raw'] = max(mis)
runtimes['specgwl-asymmetric-raw'] = sum(rt)
# avetimes['specgwl-asymmetric-raw'] = np.mean(rt)
# Noisy