-
Notifications
You must be signed in to change notification settings - Fork 6
/
Copy pathsim_end_to_end_count_based_estimates.py
256 lines (200 loc) · 9.57 KB
/
sim_end_to_end_count_based_estimates.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
# -*- coding: utf-8 -*-
"""
"CHIP end-to-end parameter Estimation"
Empirically analyzing the end-to-end consistency of the CHIP parameter estimators.
@author: Makan Arastuie
"""
import os
import copy
import pickle
import numpy as np
from os.path import join
import matplotlib.pyplot as plt
from joblib import Parallel, delayed
import generative_model_utils as utils
from dataset_utils import get_script_path
import model_fitting_utils as model_utils
from sklearn.metrics import adjusted_rand_score
from spectral_clustering import spectral_cluster
from sklearn.linear_model import LinearRegression
from parameter_estimation import estimate_hawkes_from_counts
result_file_path = join(get_script_path(), 'storage', 'results', 'end_to_end_count_based_estimate')
run_analysis = True
run_plotting = True
run_regression = True
asy_scalar = 100
# # sim params
end_time = 100 * asy_scalar
mu_diag = 0.11 / asy_scalar
mu_off_diag = 0.1 / asy_scalar
alpha_diag = 0.11
alpha_off_diag = 0.09
beta_diag = 0.14
beta_off_diag = 0.16
class_probs = [0.25, 0.25, 0.25, 0.25]
num_nodes_to_test = np.logspace(5, 8, num=7, dtype=np.int32, base=2) # 256
num_simulations = 100
n_cores = 6
n_classes = len(class_probs)
def calc_mean_and_error_of_count_estiamte(n_nodes):
params = {'number_of_nodes': n_nodes,
'class_probabilities': class_probs,
'end_time': end_time,
'mu_diag': mu_diag,
'mu_off_diag': mu_off_diag,
'alpha': alpha_off_diag,
'alpha_diag': alpha_diag,
'beta': beta_off_diag,
'beta_diag': beta_diag,
'scale': False}
event_dict, true_node_membership = utils.simulate_community_hawkes(params)
invalid_cluster = True
while invalid_cluster:
# Spectral clustering on aggregated adjacency matrix
agg_adj = utils.event_dict_to_aggregated_adjacency(n_nodes, event_dict)
node_membership = spectral_cluster(agg_adj, num_classes=n_classes, verbose=False)
unique_vals, cnts = np.unique(node_membership, return_counts=True)
invalid_cluster = len(unique_vals) != n_classes
if len(unique_vals) != n_classes:
print(unique_vals, cnts)
sc_rand = adjusted_rand_score(true_node_membership, node_membership)
sc_rand = np.zeros((n_classes, n_classes)) + sc_rand # match the shape of other params to retrieve easily
# param estimation with estimated communities
bp_mu, bp_alpha, bp_beta, bp_alpha_beta_ratio = model_utils.estimate_bp_hawkes_params(event_dict,
node_membership,
end_time,
n_classes)
# param estimation with known communities. k_ is for known_
k_bp_mu, k_bp_alpha, k_bp_beta, k_bp_alpha_beta_ratio = model_utils.estimate_bp_hawkes_params(event_dict,
true_node_membership,
end_time,
n_classes)
return bp_mu, bp_alpha_beta_ratio, bp_alpha, bp_beta, sc_rand, k_bp_mu, k_bp_alpha_beta_ratio, k_bp_alpha, k_bp_beta
true_mu = np.zeros((n_classes, n_classes)) + mu_off_diag
np.fill_diagonal(true_mu, mu_diag)
true_alpha = np.zeros((n_classes, n_classes)) + alpha_off_diag
np.fill_diagonal(true_alpha, alpha_diag)
true_beta = np.zeros((n_classes, n_classes)) + beta_off_diag
np.fill_diagonal(true_beta, beta_diag)
true_ratio = true_alpha / true_beta
params = {
'mu': ['mu_mse', 'mu_mse_err'],
'm': ['ratio_mse', 'ratio_mse_err'],
'alpha': ['alpha_mse', 'alpha_mse_err'],
'beta': ['beta_mse', 'beta_mse_err']
}
# known_communities_error
kce = {}
for param, err in params.items():
kce[err[0]] = []
kce[err[1]] = []
# estimated_communities_error
ece = copy.deepcopy(kce)
ece['rand_mean'] = []
ece['rand_mean_err'] = []
if run_analysis:
for j, n_nodes in enumerate(num_nodes_to_test):
results = Parallel(n_jobs=n_cores)(delayed(calc_mean_and_error_of_count_estiamte)
(n_nodes) for i in range(num_simulations))
results = np.asarray(results, dtype=np.float)
results = np.reshape(results, (num_simulations, 9, n_classes, n_classes))
print(f"Done simulations with {n_nodes} nodes. ARI: {np.mean(results[:, 4, 0, 0])}")
# estimated communities
mu_mse_temp = np.power(results[:, 0, :, :] - true_mu, 2)
ece['mu_mse'].append(np.mean(mu_mse_temp))
ece['mu_mse_err'].append(2 * np.std(mu_mse_temp) / np.sqrt(mu_mse_temp.size))
ratio_mse_temp = np.power(results[:, 1, :, :] - true_ratio, 2)
ece['ratio_mse'].append(np.mean(ratio_mse_temp))
ece['ratio_mse_err'].append(2 * np.std(ratio_mse_temp) / np.sqrt(ratio_mse_temp.size))
alpha_mse_temp = np.power(results[:, 2, :, :] - true_alpha, 2)
ece['alpha_mse'].append(np.mean(alpha_mse_temp))
ece['alpha_mse_err'].append(2 * np.std(alpha_mse_temp) / np.sqrt(alpha_mse_temp.size))
beta_mse_temp = np.power(results[:, 3, :, :] - true_beta, 2)
ece['beta_mse'].append(np.mean(beta_mse_temp))
ece['beta_mse_err'].append(2 * np.std(beta_mse_temp) / np.sqrt(beta_mse_temp.size))
rand_mean_temp = np.mean(results[:, 4, 0, 0])
ece['rand_mean'].append(rand_mean_temp)
ece['rand_mean_err'].append(2 * np.std(results[:, 4, 0, 0]) / np.sqrt(results[:, 4, 0, 0].size))
# known communities
mu_mse_temp = np.power(results[:, 5, :, :] - true_mu, 2)
kce['mu_mse'].append(np.mean(mu_mse_temp))
kce['mu_mse_err'].append(2 * np.std(mu_mse_temp) / np.sqrt(mu_mse_temp.size))
ratio_mse_temp = np.power(results[:, 6, :, :] - true_ratio, 2)
kce['ratio_mse'].append(np.mean(ratio_mse_temp))
kce['ratio_mse_err'].append(2 * np.std(ratio_mse_temp) / np.sqrt(ratio_mse_temp.size))
alpha_mse_temp = np.power(results[:, 7, :, :] - true_alpha, 2)
kce['alpha_mse'].append(np.mean(alpha_mse_temp))
kce['alpha_mse_err'].append(2 * np.std(alpha_mse_temp) / np.sqrt(alpha_mse_temp.size))
beta_mse_temp = np.power(results[:, 8, :, :] - true_beta, 2)
kce['beta_mse'].append(np.mean(beta_mse_temp))
kce['beta_mse_err'].append(2 * np.std(beta_mse_temp) / np.sqrt(beta_mse_temp.size))
with open(join(result_file_path, 'mses.pckl'), 'wb') as handle:
pickle.dump([ece, kce], handle, protocol=pickle.HIGHEST_PROTOCOL)
with open(join(result_file_path, 'mses.pckl'), 'rb') as handle:
ece, kce = pickle.load(handle)
print("Estimated communities:")
print('Rand mean:')
print(ece['rand_mean'])
for param, err in params.items():
print(param, "MSE:")
print(ece[err[0]])
print("\nKnown communities:")
for param, err in params.items():
print(param, "MSE:")
print(kce[err[0]])
if run_plotting:
# rand index for estimated communities
plt.ion()
plt.subplots(figsize=(3.8, 3))
plt.bar(range(len(num_nodes_to_test)), ece['rand_mean'], yerr=ece['rand_mean_err'])
plt.xlabel("Number of Nodes", fontsize=16)
plt.ylabel("Mean-squared Error", fontsize=16)
plt.xticks(range(len(num_nodes_to_test)), num_nodes_to_test)
plt.tick_params(labelsize=12)
plt.tight_layout()
plt.savefig(join(result_file_path, 'plots', 'estimated_consistent_rand_mean.pdf'))
for param, err in params.items():
# estimated communities
mse, mse_err = err[0], err[1]
plt.ion()
plt.subplots(figsize=(3.8, 3))
plt.bar(range(len(num_nodes_to_test)), ece[mse], yerr=ece[mse_err], log=True)
plt.xlabel("Number of Nodes", fontsize=16)
plt.ylabel("Mean-squared Error", fontsize=16)
plt.xticks(range(len(num_nodes_to_test)), num_nodes_to_test)
plt.tick_params(labelsize=12)
plt.tight_layout()
plt.savefig(join(result_file_path, 'plots', f'estimated_consistent_{param}_mse.pdf'))
# known communities
plt.ion()
plt.subplots(figsize=(3.8, 3))
plt.bar(range(len(num_nodes_to_test)), kce[mse], yerr=kce[mse_err], log=True)
plt.xlabel("Number of Nodes", fontsize=16)
plt.ylabel("Mean-squared Error", fontsize=16)
plt.xticks(range(len(num_nodes_to_test)), num_nodes_to_test)
plt.tick_params(labelsize=12)
plt.tight_layout()
plt.savefig(join(result_file_path, 'plots', f'known_consistent_{param}_mse.pdf'))
if run_regression:
print("\nRegression: \n")
print("Estimated Communities:\n")
start_idx = 0
end_idx = 3
num_nodes = num_nodes_to_test[start_idx:end_idx]
for param, err in params.items():
print(param, '(estimated communities)')
x = np.log(num_nodes).reshape(len(num_nodes), 1)
y = np.log(ece[err[0]][start_idx:end_idx]).reshape(len(ece[err[0]][start_idx:end_idx]), 1)
reg = LinearRegression().fit(x, y)
print("R^2:", reg.score(x, y))
print("coef:", reg.coef_)
print('intercept:', reg.intercept_)
print()
print(param, '(known communities)')
x = np.log(num_nodes).reshape(len(num_nodes), 1)
y = np.log(kce[err[0]][start_idx:end_idx]).reshape(len(kce[err[0]][start_idx:end_idx]), 1)
reg = LinearRegression().fit(x, y)
print("R^2:", reg.score(x, y))
print("coef:", reg.coef_)
print('intercept:', reg.intercept_)
print()