-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathcompute_activities_newest.py
186 lines (136 loc) · 7.55 KB
/
compute_activities_newest.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
from __future__ import print_function
import torch
import numpy as np
import scipy.io
import math
from os import listdir
import matplotlib.pyplot as plt
import os
import torch
import argparse
from scipy.stats import ks_2samp
parser = argparse.ArgumentParser()
parser.add_argument("--vid_path", type=str, default='/home/dvoina/simple_vids/moving_videos_bsr_jumps_simple_3px_valid',
required=False, help="Set path for image dataset")
parser.add_argument("--Nf2", type=int, default=5, required=False, help="number of VIP neurons/u.s.")
parser.add_argument("--tau", type=int, default=2, required=False, help="tau/synaptic delay")
args = parser.parse_args()
vid_path = args.vid_path
tau = args.tau
fr_pervid = 49 # how many frames are in a video after convolution w/ filters has been done!
Nf2 = args.Nf2
XTrain = np.load('/home/dvoina/simple_vids/relevant/XTrain_forPython_34filters_noNoise_reviewSimple.npy')
address = vid_path
folder_list = listdir(address)
no_folders = np.shape(folder_list)[0]
W_stat = np.load('/home/dvoina/simple_vids/results/W_43x43_34filters_static_simple_3px_ReviewSparse2.npy')
W_mov = np.load('/home/dvoina/simple_vids/results/W_43x43_34filters_moving_simple_3px_tau' + str(tau) + '_ReviewSparse.npy')
Wp2v_load = np.load('/home/dvoina/simple_vids/results/Wp2v_34_param_simple3px_sst2vip_Nf' + str(Nf2) + '_review2.npy')
Wv2p_load = np.load('/home/dvoina/simple_vids/results/Wv2p_34_param_simple3px_sst2vip_Nf' + str(Nf2) + '_review2.npy')
Wv2s_load = np.load('/home/dvoina/simple_vids/results/Wv2s_34_param_simple3px_sst2vip_Nf' + str(Nf2) + '_review2.npy')
Wp2v = Wp2v_load[-3, 2, :, :, :, :]
Wv2p = -Wv2p_load[-3, 2, :, :, :, :]
Wv2s = -Wv2s_load[-3, 2, :, :, :, :]
#Wp2v = Wp2v_load[-1, -3, :, :, :, :]
#Wv2p = -Wv2p_load[-1, -3, :, :, :, :]
#Wv2s = -Wv2s_load[-1, -3, :, :, :, :]
W_s2p = np.copy(W_stat)
W_s2p[W_s2p > 0] = 0
W_moving_minus_static = W_mov - W_stat
W_stat_matrix_torch = torch.from_numpy(W_stat).float()
W_mov_matrix_torch = torch.from_numpy(W_mov).float()
W_s2p_torch = torch.from_numpy(W_s2p).float()
W_moving_minus_static_torch = torch.from_numpy(W_moving_minus_static).float()
Wp2v_torch = torch.from_numpy(Wp2v).float()
Wv2p_torch = torch.from_numpy(Wv2p).float()
Wv2s_torch = torch.from_numpy(Wv2s).float()
folder_list = listdir(address)
no_folders = np.shape(folder_list)[0]
input_gaussian = torch.load("/home/dvoina/simple_vids/input_gaussian.npy")
input_constant = torch.load("/home/dvoina/simple_vids/input_constant.npy")
bar_H = torch.load("/home/dvoina/simple_vids/bar_H.npy")
bar_V = torch.load("/home/dvoina/simple_vids/bar_V.npy")
bar_iso_H = torch.load("/home/dvoina/simple_vids/bar_iso_H.npy")
bar_iso_V = torch.load("/home/dvoina/simple_vids/bar_iso_V.npy")
bar_cross_HV = torch.load("/home/dvoina/simple_vids/bar_cross_HV.npy")
bar_cross_VH = torch.load("/home/dvoina/simple_vids/bar_cross_VH.npy")
input_gaussian = torch.load("/home/dvoina/simple_vids/input_gaussian_static.npy")
input_constant = torch.load("/home/dvoina/simple_vids/input_constant_static.npy")
print("compute Activities static and Activities moving")
#no_folders = 1
A_static = np.zeros((no_folders,1,47,34,125,125))
A_moving = np.zeros((no_folders,1,47,34,125,125))
A_approx = np.zeros((no_folders,47,34,121,121))
for i in range(no_folders):
print(i)
#XTrain_main = XTrain[i*fr_pervid+2:i*fr_pervid+fr_pervid,0,:,:,:]
#XTrain_prev = XTrain[i*fr_pervid:i*fr_pervid+fr_pervid-2,0,:,:,:]
XTrain_main = XTrain[i*fr_pervid+tau:i*fr_pervid+fr_pervid, 0, :, :, :]
XTrain_prev = XTrain[i*fr_pervid:i*fr_pervid+fr_pervid-tau, 0, :, :, :]
#XTrain_main = input_gaussian[i*fr_pervid+tau:i*fr_pervid+fr_pervid,:,:]
#XTrain_prev = input_gaussian[i*fr_pervid:i*fr_pervid+fr_pervid-tau,:,:]
#XTrain_main = input_constant[i,:,:,:,:]
#XTrain_prev = input_constant[i,:,:,:,:]
#XTrain_main = bar_cross_VH
#XTrain_prev = bar_cross_VH
XTrain_main = torch.from_numpy(XTrain_main).squeeze(1)
XTrain_prev = torch.from_numpy(XTrain_prev).squeeze(1)
y_mov = torch.nn.functional.conv2d(XTrain_prev, W_mov_matrix_torch)
y_stat = torch.nn.functional.conv2d(XTrain_main, W_stat_matrix_torch)
A_moving2 = XTrain_main[:, :, 21:-21, 21:-21] * (
torch.ones(XTrain_main.size()[0], XTrain_main.size()[1], XTrain_main.size()[2] - 42,
XTrain_main.size()[3] - 42) + y_mov)
A_static2 = XTrain_main[:, :, 21:-21, 21:-21] * (
torch.ones(XTrain_main.size()[0], XTrain_main.size()[1], XTrain_main.size()[2] - 42,
XTrain_main.size()[3] - 42) + y_stat)
A_moving2 = A_moving2.numpy()
A_static2 = A_static2.numpy()
A_vip = torch.nn.functional.conv2d(XTrain_main, Wp2v_torch)
A_moving2 = A_moving2[np.newaxis, :, :, :, :]
A_static2 = A_static2[np.newaxis, :, :, :, :]
Activities2 = np.concatenate((A_static2, A_moving2), axis=0)
A_static[i,:,:,:,:,:] = A_static2
A_moving[i,:,:,:,:,:] = A_moving2
np.save('/home/dvoina/simple_vids/Activities_movstat_34filters_simple_tau' + str(tau) + "/"+ folder_list[i] + '_activities_noNoise.npy', Activities2)
#np.save('/home/dvoina/simple_vids/XTrain_main_34filters_simple_tau' + str(tau) + '/'+ folder_list[i] + '_xmain_noNoise.npy', XTrain_main)
np.save('/home/dvoina/simple_vids/Activities_vip_34_noNoise_new_from_fc/' + folder_list[i] + '_activities.npy', A_vip)
del A_moving2, A_static2, Activities2, y_mov, y_stat, XTrain_main, XTrain_prev
print(np.mean(A_static))
print(np.mean(A_moving))
print(ks_2samp(A_static.flatten(), A_moving.flatten()))
print("compute approx Activities and Activities of VIP/SST and VIP/SST contributions")
for i in range(no_folders):
print(i)
XTrain_main = XTrain[i * fr_pervid + tau:i * fr_pervid + fr_pervid, 0, :, :, :]
XTrain_prev = XTrain[i * fr_pervid:i * fr_pervid + fr_pervid - tau, 0, :, :, :]
#XTrain_main = input_gaussian[i*fr_pervid+tau:i*fr_pervid+fr_pervid,:,:]
#XTrain_prev = input_gaussian[i*fr_pervid:i*fr_pervid+fr_pervid-tau,:,:]
#XTrain_main = bar_H
#XTrain_prev = bar_H
XTrain_main = torch.from_numpy(XTrain_main).squeeze(1)
XTrain_prev = torch.from_numpy(XTrain_prev).squeeze(1)
y_stat = torch.nn.functional.conv2d(XTrain_prev, W_stat_matrix_torch)
h1 = torch.nn.functional.conv2d(XTrain_prev, Wp2v_torch)
x1 = torch.nn.functional.conv2d(h1, Wv2p_torch)
h2 = torch.nn.functional.conv2d(h1, Wv2s_torch)
x2 = torch.nn.functional.conv2d(h2, W_s2p_torch)
x1 = x1[:, :, 21:-21, 21:-21]
A_approx2 = XTrain_main[:, :, 23:-23, 23:-23] * (
torch.ones(XTrain_main.size()[0], XTrain_main.size()[1], XTrain_main.size()[2] - 46,
XTrain_main.size()[3] - 46) + y_stat[:, :, 2:-2, 2:-2] + x1 + x2)
A_approx[i,:,:,:,:] = A_approx2
"""
A_vip = torch.nn.functional.conv2d(A_approx2, Wp2v_torch)
C_vip = torch.nn.functional.conv2d(A_vip, Wv2p_torch)
A_sst = XTrain_main[:, :, 25:-25, 25:-25] + torch.nn.functional.conv2d(
torch.nn.functional.conv2d(A_approx2, Wp2v_torch), Wv2s_torch)
C_sst = torch.nn.functional.conv2d(A_sst, W_s2p_torch)
"""
A_vip1 = torch.nn.functional.conv2d(A_approx2, Wp2v_torch)
A_sst1 = XTrain_main[:, :, 25:-25, 25:-25] + torch.nn.functional.conv2d(
torch.nn.functional.conv2d(A_approx2, Wp2v_torch), Wv2s_torch)
np.save('/home/dvoina/simple_vids/Activities_vip_34_noNoise_new/' + folder_list[i] + '_activities.npy',
A_vip1)
np.save('/home/dvoina/simple_vids/Activities_approx_34_noNoise_new/' + folder_list[i] + '_activities_noNoise.npy',
A_approx2.numpy())
del A_approx2, XTrain_main, XTrain_prev, A_vip1