/
create_accuracy_figure.py
186 lines (175 loc) · 8.58 KB
/
create_accuracy_figure.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
# Code for
# Physics solutions for machine learning privacy leaks
# arXiv:2202.12319
#
# Authors: Alejandro Pozas-Kerstjens and Senaida Hernandez-Santana
#
# Requires: glob, os for filesystem operations
# matplotlib for plotting
# numpy for array operations
# pandas for dataset operations
# pickle for data loading
# seaborn for plot visuals
# tensorflow, torch for ML
# tqdm for progress bar
# Last modified: Feb, 2023
###############################################################################
# This file generates Figure 2c in the paper, which compares the accuracy of
# neural networks and MPS trained to predict the outcome of a COVID-19
# infection given demographics and symptoms.
###############################################################################
import glob, os
import matplotlib as mpl
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import pickle
import seaborn as sns
import tensorflow as tf
import torch
from matrix_product_states.utils_mps import preprocess_mps
from matrix_product_states.classifier import MatrixProductState
from neural_networks.utils_nn import SimpleNNModel, preprocess_nn
from tqdm import tqdm
###############################################################################
# Style data
###############################################################################
sns.set_theme()
sns.set(style='whitegrid', font_scale=1.8)
mpl.rcParams['lines.linewidth'] = 0.75
mpl.rcParams['errorbar.capsize'] = 3
mpl.rcParams.update({'font.size': 18})
plt.rc('text', usetex=True)
plt.rc('text.latex', preamble=r'\usepackage{amsmath}')
plt.rc('font', family='serif')
###############################################################################
# Common information
###############################################################################
imbalance_levels = np.array([0.5, 0.51, 0.55, 0.6, 0.7, 0.8, 0.9, 0.95, 0.99])
data_dir = 'datasets'
models_dir = 'models'
results_dir = 'results'
all_data = f'{data_dir}/covid_argentina_colombia_until20210322.csv'
df = pd.read_csv(all_data, index_col=0)
device = torch.device('cpu')
###############################################################################
# Neural network data
###############################################################################
nn_data_exists = (len([file for file in glob.glob(f'{results_dir}/*_nn_*')])
== 4)
if not nn_data_exists:
print('Calculating data for neural networks')
df_nn = df.copy()
df_nn = preprocess_nn(df_nn)
labels = df_nn['recovered'].to_numpy()
feature_dataset = df_nn.drop(['recovered'], axis=1)
data = feature_dataset.to_numpy()
data_tensor = torch.from_numpy(data).type(torch.FloatTensor).to(device)
data_labels = torch.from_numpy(labels).type(torch.LongTensor).to(device)
NN = SimpleNNModel().to(device)
all_accuracies = []
for dataset in tqdm(os.listdir(f'{models_dir}/nn')):
if len(dataset.split('.')) == 1: # Discard non-folders
accuracies_odd = [[], [], [], [], [], [], [], [], []]
accuracies_even = [[], [], [], [], [], [], [], [], []]
accuracies = [accuracies_odd, accuracies_even]
for model in os.listdir(f'{models_dir}/nn/{dataset}'):
if len(model.split('.')[-1]) == 2: # Take .pt models
try:
imbalance = float(model.split('_')[1][:4])
except ValueError:
imbalance = float(model.split('_')[1][:3])
imbalance_pos = np.where(
imbalance_levels == imbalance
)[0].item()
typ = 0 if model.split('_')[1][-3:] == 'odd' else 1
NN.load_state_dict(torch.load(
f'{models_dir}/nn/{dataset}/{model}'))
acc = (NN(data_tensor).argmax(dim=1)
== data_labels).sum().item()
acc /= len(data_labels)
accuracies[typ][imbalance_pos].append(acc)
all_accuracies.append(accuracies)
all_accuracies = np.array(all_accuracies)
avg_accuracies_per_dataset = all_accuracies.mean(-1)
acc_nns_odd = avg_accuracies_per_dataset[:, 0, :].mean(0)
std_nns_odd = avg_accuracies_per_dataset[:, 0, :].std(0)
acc_nns_even = avg_accuracies_per_dataset[:, 1, :].mean(0)
std_nns_even = avg_accuracies_per_dataset[:, 1, :].std(0)
np.savetxt(f'{results_dir}/acc_nn_odd.txt', acc_nns_odd)
np.savetxt(f'{results_dir}/acc_nn_even.txt', acc_nns_even)
np.savetxt(f'{results_dir}/std_nn_odd.txt', std_nns_odd)
np.savetxt(f'{results_dir}/std_nn_even.txt', std_nns_even)
else:
print('Loading data for neural networks')
acc_nns_odd = np.loadtxt(f'{results_dir}/acc_nn_odd.txt')
acc_nns_even = np.loadtxt(f'{results_dir}/acc_nn_even.txt')
std_nns_odd = np.loadtxt(f'{results_dir}/std_nn_odd.txt')
std_nns_even = np.loadtxt(f'{results_dir}/std_nn_even.txt')
###############################################################################
# MPS data
###############################################################################
mps_data_exists = (len([file
for file in glob.glob(f'{results_dir}/*_mps_*.txt')])
== 4)
if not mps_data_exists:
print('Calculating data for MPS')
MPS = MatrixProductState(5, 2, 2, 2)
data_tensor, data_labels = preprocess_mps(df)
data_tensor = tf.cast(data_tensor, tf.float32)
data_labels = tf.cast(data_labels, tf.float32)
all_accuracies = []
for dataset in tqdm(os.listdir(f'{models_dir}/mps')):
if len(dataset.split('.')) == 1: # Discard non-folders
accuracies_odd = [[], [], [], [], [], [], [], [], []]
accuracies_even = [[], [], [], [], [], [], [], [], []]
accuracies = [accuracies_odd, accuracies_even]
for model in os.listdir(f'{models_dir}/mps/{dataset}'):
if len(model.split('.')[-1]) == 6: # Take .pickle models
try:
imbalance = float(model.split('_')[1][:4])
except ValueError:
imbalance = float(model.split('_')[1][:3])
imbalance_pos = np.where(
imbalance_levels == imbalance
)[0].item()
typ = 0 if model.split('_')[1][-3:] == 'odd' else 1
with open(f'{models_dir}/mps/{dataset}/{model}', 'rb') as f:
params = pickle.load(f)
MPS.load_numpy(params)
acc = MPS.accuracy(data_tensor, data_labels)
accuracies[typ][imbalance_pos].append(acc)
all_accuracies.append(accuracies)
all_accuracies = np.array(all_accuracies)
avg_accuracies_per_dataset = all_accuracies.mean(-1)
acc_mps_odd = avg_accuracies_per_dataset[:, 0, :].mean(0)
std_mps_odd = avg_accuracies_per_dataset[:, 0, :].std(0)
acc_mps_even = avg_accuracies_per_dataset[:, 1, :].mean(0)
std_mps_even = avg_accuracies_per_dataset[:, 1, :].std(0)
np.savetxt(f'{results_dir}/acc_mps_odd.txt', acc_mps_odd)
np.savetxt(f'{results_dir}/acc_mps_even.txt', acc_mps_even)
np.savetxt(f'{results_dir}/std_mps_odd.txt', std_mps_odd)
np.savetxt(f'{results_dir}/std_mps_even.txt', std_mps_even)
else:
print('Loading data for MPS')
acc_mps_odd = np.loadtxt(f'{results_dir}/acc_mps_odd.txt')
acc_mps_even = np.loadtxt(f'{results_dir}/acc_mps_even.txt')
std_mps_odd = np.loadtxt(f'{results_dir}/std_mps_odd.txt')
std_mps_even = np.loadtxt(f'{results_dir}/std_mps_even.txt')
###############################################################################
# Plot
###############################################################################
print(acc_mps_odd)
print(acc_mps_even)
plt.errorbar(imbalance_levels, acc_mps_odd, std_mps_odd,
marker='o', label='MPS, odd')
plt.errorbar(imbalance_levels, acc_mps_even, std_mps_even,
marker='o', label='MPS, even')
plt.errorbar(imbalance_levels, acc_nns_odd, std_nns_odd,
marker='*', markersize=10, label='NNs, odd')
plt.errorbar(imbalance_levels, acc_nns_even, std_nns_even,
marker='*', markersize=10, label='NNs, even')
plt.legend(loc=3, prop={'size': 14})
plt.xlabel('Percentage of majority irrelevant class')
plt.ylabel('Model accuracy')
plt.savefig('models.pdf', bbox_inches='tight')