def genoptfunction(graph, samples=1000):
live_graphs = sample_live_icm(graph, samples)
f_multi = make_multilinear_objective_samples(live_graphs,
list(graph.nodes()),
list(graph.nodes()),
np.ones(len(graph)))
f_set = multi_to_set(f_multi, graph)
return f_set
gr_values[graphname][attribute] = np.zeros((num_runs, nvalues))
group_size[graphname][attribute] = np.zeros((num_runs, nvalues))
for alg in algorithms:
alg_values[alg][graphname][attribute] = np.zeros((num_runs, nvalues))
alg_seeds[alg][graphname][attribute] = [[] for _ in range(num_runs)]
fair_vals_attr = np.zeros((num_runs, len(attributes)))
greedy_vals_attr = np.zeros((num_runs, len(attributes)))
pof = np.zeros((num_runs, len(attributes)))
include_total = False
for run in range(num_runs):
print(graphname, run)
live_graphs = sample_live_icm(g, 1000)
group_indicator = np.ones((len(g.nodes()), 1))
val_oracle = make_multilinear_objective_samples_group(live_graphs, group_indicator, list(g.nodes()), list(g.nodes()), np.ones(len(g)))
grad_oracle = make_multilinear_gradient_group(live_graphs, group_indicator, list(g.nodes()), list(g.nodes()), np.ones(len(g)))
def f_multi(x):
return val_oracle(x, 1000).sum()
#f_multi = make_multilinear_objective_samples(live_graphs, list(g.nodes()), list(g.nodes()), np.ones(len(g)))
f_set = multi_to_set(f_multi)
#find overall optimal solution
S, obj = greedy(list(range(len(g))), budget, f_set)