-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathGLM_regression_analysis.py
229 lines (174 loc) · 7.07 KB
/
GLM_regression_analysis.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
import numpy as np
import torch
import torch.nn as nn
from torch.autograd import Variable
import torch.nn.functional as F
import os
import matplotlib.pyplot as plt
from scipy.io import loadmat
import pywt
import pywt.data
from scipy.stats import pearsonr
import pylab
from scipy.io import loadmat
import scipy.stats
import statsmodels.api as sm
A_vip = []
A_stat = []
A_mov = []
A_approx = []
f2_array = []
f3_array = []
f4_array = []
files_list = os.listdir("/home/dvoina/simple_vids/Activities_vip_34_noNoise_new_from_fc")
vid_path = '/home/dvoina/simple_vids/moving_videos_bsr_jumps_simple_3px_valid'
avgpool = torch.nn.AvgPool2d(5, stride=9, padding=0)
for j, file in enumerate(files_list):
print(j)
a = np.load("/home/dvoina/simple_vids/Activities_vip_34_noNoise_new_from_fc/" + file)
a2 = a
# /np.resize(np.max(np.abs(a), axis=1), (47,1,165,165))
A_vip.append(a2)
# files_list = os.listdir("Activities_movstat_34filters_simple_tau2")
files_list = os.listdir("/home/dvoina/simple_vids/Activities_approx_34_noNoise_new")
for j, file in enumerate(files_list):
print(j)
A = np.load("/home/dvoina/simple_vids/Activities_approx_34_noNoise_new/" + file)
A2 = A / np.resize(np.max(np.abs(A), axis=1), (47, 1, 121, 121))
A_approx.append(A2)
# A_stat.append(A[0,:,:,:,:])
# A_mov.append(A[1,:,:,:,:])
files_list = os.listdir("/home/dvoina/simple_vids/Activities_movstat_34filters_simple_tau2")
for j, file in enumerate(files_list):
print(j)
A = np.load("/home/dvoina/simple_vids/Activities_movstat_34filters_simple_tau2/" + file)
print(A.shape)
A_stat_ = A[0, :, :, :, :] / np.resize(np.max(np.abs(A[0, :, :, :, :]), axis=1), (47, 1, 125, 125))
A_mov_ = A[1, :, :, :, :] / np.resize(np.max(np.abs(A[1, :, :, :, :]), axis=1), (47, 1, 125, 125))
A_stat.append(A_stat_)
A_mov.append(A_mov_)
files_list = os.listdir("/home/dvoina/simple_vids/moving_videos_bsr_jumps_simple_3px_valid")
for j, file in enumerate(files_list):
print(file)
vid = loadmat(vid_path + '/' + file)
for frame in range(47):
coeffs2 = pywt.dwt2(vid["s_modified"][3 + frame, :, :], 'db4')
f1, (f2, f3, f4) = coeffs2
f2_avg = avgpool(torch.from_numpy(f2).unsqueeze(0)).squeeze()
f3_avg = avgpool(torch.from_numpy(f3).unsqueeze(0)).squeeze()
f4_avg = avgpool(torch.from_numpy(f4).unsqueeze(0)).squeeze()
f2_array.append(f2_avg)
f3_array.append(f3_avg)
f4_array.append(f4_avg)
f2_array = torch.stack(f2_array)
f3_array = torch.stack(f3_array)
f4_array = torch.stack(f4_array)
f2_array = np.resize(np.array(f2_array), (4700, 10*10))
f3_array = np.resize(np.array(f3_array), (4700, 10*10))
f4_array = np.resize(np.array(f4_array), (4700, 10*10))
features_array = np.concatenate((f2_array, f3_array, f4_array), axis=1)
A_vip = np.array(A_vip)
A_vip = np.resize(A_vip, (4700, 5*165*165))
A_stat = np.array(A_stat)
A_stat = np.resize(A_stat, (4700, 34*125*125))
A_mov = np.array(A_mov)
A_mov = np.resize(A_mov, (4700, 34*125*125))
A_approx = np.array(A_approx)
A_approx = np.resize(A_approx, (4700,34,121,121))
"""
results = np.zeros((5,123, 123, 300))
ci_results = np.zeros((5,123,123, 300, 2))
pval_results = np.zeros((5,123,123, 300))
x = 0
for i in range(5):
for s1 in range(x, x + 12):
for s2 in range(x, x + 12):
print(i, s1, s2)
glm_binom = sm.GLM(np.squeeze(A_vip[:, i, s1, s2]), features_array, family=sm.families.Binomial())
res = glm_binom.fit()
results[i, s1 - x, s2 - x, :] = res.params
ci_results[i, s1 - x, s2 - x, :, :] = res.conf_int(alpha=0.05, cols=None)
pval_results[i, s1 - x, s2 - x, :] = res.pvalues
x = 0
results2 = np.zeros((34, 121, 121, 300))
ci_results2 = np.zeros((34, 121, 121, 300, 2))
pval_results2 = np.zeros((34, 121, 121, 300))
for i in range(34):
for s1 in range(x + 2, x + 12):
for s2 in range(x + 2, x + 12):
print(i, s1, s2)
glm_binom = sm.GLM(np.squeeze(A_approx[:, i, s1, s2]), features_array, family=sm.families.Binomial())
# family=sm.families.binomial)
res = glm_binom.fit()
results2[i, s1 - 2 - x, s2 - 2 - x, :] = res.params
ci_results2[i, s1 - 2 - x, s2 - 2 - x, :, :] = res.conf_int(alpha=0.05, cols=None)
pval_results2[i, s1 - 2 - x, s2 - 2 - x, :] = res.pvalues
results_flat = np.reshape(results[:,1:11,1:11,:], (5*10*10, 300))
results2_flat = np.reshape(results2[:,0:10,0:10,:], (34*10*10, 300))
corr_coeff_regression = results_flat.dot(results2_flat.T) #vip vs pyr
"""
results = np.zeros((10000, 300))
ci_results = np.zeros((10000, 300, 2))
pval_results = np.zeros((10000, 300))
start = 0
end = start + 10000
for i in range(start, end + 1):
print(i)
glm_binom = sm.GLM(A_vip[:, i], features_array, family=sm.families.Poisson())
res = glm_binom.fit()
results[i - start, :] = res.params
ci_results[i - start, :, :] = res.conf_int(alpha=0.05, cols=None)
pval_results[i - start, :] = res.pvalues
plt.figure()
plt.plot(range(10000), np.mean(results[0:10000,0:100], axis=1))
plt.xlabel("neurons")
plt.savefig("avg_neurons_vs_horiz.pdf", format="pdf", bbox_inches="tight")
#plt.plot(range(10000), np.mean(ci_results[0:10000,0:100,0], axis=1))
#plt.plot(range(10000), np.mean(ci_results[0:10000,0:100,1], axis=1))
plt.figure()
plt.plot(range(10000), np.mean(results[0:10000,100:200], axis=1))
plt.xlabel("neurons")
plt.savefig("avg_neurons_vs_vert.pdf", format="pdf", bbox_inches="tight")
plt.figure()
plt.plot(range(10000), np.mean(results[0:10000,200:300], axis=1))
plt.xlabel("neurons")
print(np.mean(np.abs(results[:,0:100])))
print(np.mean(np.abs(results[:,100:200])))
print(np.mean(np.abs(results[:,200:300])))
res1 = results[:,0:100]
res2 = results[:,0:200]
res3 = results[:,0:300]
res1 = res1.flatten()
res1_pos = res1[res1>=0]
res1_neg = res1[res1<=0]
print("res1", np.mean(res1_pos), np.mean(res1_neg))
res2 = res2.flatten()
res2_pos = res2[res2>=0]
res2_neg = res2[res2<=0]
print("res2", np.mean(res2_pos), np.mean(res2_neg))
res3 = res3.flatten()
res3_pos = res3[res3>=0]
res3_neg = res3[res3<=0]
print("res3", np.mean(res3_pos), np.mean(res3_neg))
print(np.mean(results[:,0:100]))
print(np.mean(results[:,100:200]))
print(np.mean(results[:,200:300]))
plt.bar([0, 1, 2], [np.mean(results[:,0:100]), np.mean(results[:,100:200]), np.mean(results[:,200:300])])
plt.savefig("compare_horiz_vert.pdf", format="pdf", bbox_inches="tight")
mean_f1 = np.mean(results[:,0:100], axis=0)
mean_f2 = np.mean(results[:,100:200], axis=0)
mean_f3 = np.mean(results[:,200:300], axis=0)
#plt.figure()
plt.plot(range(100), mean_f1)
plt.xlabel("features")
#plt.figure()
plt.plot(range(100), mean_f2)
plt.xlabel("features")
#plt.figure()
plt.plot(range(100), mean_f3)
plt.xlabel("features")
plt.legend(["horizontal", "vertical", "diagonal"])
plt.savefig("compare_features_vs_H_V_D.pdf", format="pdf", bbox_inches="tight")
plt.scatter(np.mean(results[:,0:100], axis=1), np.mean(results[:,100:200], axis=1))
plt.plot(np.arange(-0.15,0.5, 0.01), np.arange(-0.15,0.5, 0.01))
plt.savefig("scatter_plot_H_vs_V.pdf", format="pdf", bbox_inches="tight")