def main():
results = []
for i, p in enumerate(parameters_list):
f = NAME_TO_FUNC[p["func"]]
agents = f(**{k: v for k, v in p.items() if k != "func"})
optimal = opt.budget_constraint_min_d1(agents)
plot3d(agents, optimal_point=optimal, fname=f"figs/dataset_{i}.html")
for _ in tqdm(range(N_SAMPLES)):
agents = f(**{k: v for k, v in p.items() if k != "func"})
optimal = opt.budget_constraint_min_d1(agents)
mean = opt.mean(agents)
this_row = copy.deepcopy(p)
this_row["dataset_num"] = i
this_row["title"] = get_title_from_params(p)
this_row["optimal"] = optimal
this_row["mean"] = mean
this_row["optimal_sc"] = social_cost(agents, optimal, metrics.d_1)
this_row["mean_sc"] = social_cost(agents, mean, metrics.d_1)
this_row[
"distortion"] = this_row["mean_sc"] / this_row["optimal_sc"]
results.append(this_row)
df = pd.DataFrame(results)
df.to_csv("results/mean_approx.csv", index=False)
def plot_datasets():
for i, agents, name in enumerate(DATASETS):
optimal = opt.budget_constraint_min_d1(agents)
plot3d(agents,
optimal_point=optimal,
fname=f"figs/dataset_sequential_{i}.html")
print(distortion(agents, opt.mean(agents), optimal, metrics.d_1))
def main():
rows = []
for i, (agents, name) in enumerate(DATASETS):
results = sequential_deliberation_many_times(
agents, bargain.mean, metrics.d_1, NUM_ITERS,
NUM_SAMPLES) # list of results
optimal = opt.budget_constraint_min_d1(agents)
for r in results:
rows.append([i, name, distortion(agents, r, optimal, metrics.d_1)])
df = pd.DataFrame(rows, columns=['dataset_ind', 'title', 'distortion'])
df.to_csv("results/sequential.csv", index=False)
def main():
agents = get_d_dimensional_agents_summing_to_1_with_clusters(
NUM_AGENTS, d=DIM, noise=0.04, n_clusters=2, weights=[0.2, 0.8]
)
optimal = opt.budget_constraint_min_d1(agents)
intermediates = sequential_deliberation_list(
agents, bargain.mean, metrics.d_1, T=NUM_ITERS
)
distortions = [distortion(agents, i, optimal, metrics.d_1) for i in intermediates]
px.line(distortions).write_html("figs/line.html")
def main():
results = []
for i, p in enumerate(parameters_list):
for _ in range(100):
agents = get_d_dimensional_agents_summing_to_1_with_clusters(
**{k: v
for k, v in p.items() if k != "func"})
optimal = opt.budget_constraint_min_d1(agents)
dist = distortion(agents, opt.mean(agents), optimal, metrics.d_1)
print(dist)
if dist > 1.4:
print(agents)
print(optimal)
print(dist)
plot3d(agents,
optimal_point=optimal,
fname=f"figs/high_distortion_agents.html")
return
def sequential_deliberation_once():
agents = get_d_dimensional_agents_summing_to_1_with_clusters(
NUM_AGENTS, d=DIM, noise=0.04, n_clusters=2, weights=[0.2, 0.8])
optimal = opt.budget_constraint_min_d1(agents)
a_T = sequential_deliberation(agents,
bargain.mean,
metrics.d_1,
T=NUM_ITERS)
sc_opt = social_cost(agents, optimal, metrics.d_1)
sc_mean = social_cost(agents, opt.mean(agents), metrics.d_1)
sc_seq = social_cost(agents, a_T, metrics.d_1)
print("Social cost of the optimal: ", sc_opt)
print("Social cost of the mean: ", sc_mean)
print("Social cost of the sequential mean: ", sc_seq)
plot3d(agents, found_point=a_T, optimal_point=optimal)