-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathICA.py
448 lines (330 loc) · 18.6 KB
/
ICA.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
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
import sys, getopt
import numpy as np
import tensorflow.compat.v1 as tf
tf.disable_v2_behavior()
import time
import os
import ICA_support_lib as sup
import ICA_astrocyte as astr_astro
import ICA_ising as astr_ising
class astro_pp_model:
def __init__(self, num_of_astros=1):
self.main_Path = os.getcwd()
self.ising_model_Path = self.main_Path + '/Ising_models/'
self.data_Path = self.main_Path + '/dataFiles/'
sup.check_create_save_dir(self.data_Path)
self.astro_num = num_of_astros
def open_ising_network(self, model):
# OPENS ISING MODEL DATA
filename = 'Model_'+str(model)+'/ISING_Model_' + str(model)
names, data = sup.unpack_file(filename,self.ising_model_Path)
ind1 = data[names.index('ind1')]
ind2 = data[names.index('ind2')]
main_spins = data[names.index('main_spins')]
param_T = data[names.index('param_T')]
J = data[names.index('J')]
spin_dist = data[names.index('spin_dist')]
out_rescaled = data[names.index('out_rescaled')]
return [ind1, ind2, main_spins, param_T, J, spin_dist, out_rescaled]
############# VERSION FOR MULTI RUN, CAN FEED CUSTOM T_SCHE IN
def test_ising_w_patterns_w_astro_v5_5(self, model, save_ver, t_schd_in, ms_per_temp=10000, ini_dur_ms=60000, t_slots=18,
max_rate=20, v_beta_scale_factor=[3.0, 3.0], randomize_glut_amount=False,
glut_level=[1.0, 1.0]):
assert(len(glut_level) == len(v_beta_scale_factor) == self.astro_num)
# -SEEDSET- SET RANDOM SEED
np.random.seed()
rnd_sd = np.random.randint(0, 1000)
np.random.seed(rnd_sd)
######################## SET SEED MANUALLY ######################
# np.random.seed(121)
######################## SET SEED MANUALLY ######################
dataPath_ver = self.data_Path + 'ver_' + str(save_ver)
sup.check_create_save_dir(dataPath_ver)
print(dataPath_ver)
### START -- OPEN ISING MODEL ###
md = self.open_ising_network(model)
shp = np.shape(md[2])
isi = astr_ising.astro_pp_model_ising(synaptic_matrix_size=shp)
ind1, ind2, main_spins, param_T, J, spin_dist, feed_temp_scalar, spin_feeder = isi.reinitialize_vars(
ind1_list=md[0], ind2_list=md[1], main_spins_arr=md[2], param_T_arr=md[3], J_arr=md[4], spin_dist_arr=md[5],
out_rescaled_arr=md[6])
# ISING UPDATE OPS
energy_cur, energy_fin, energy_net = isi.pre_update_computations(main_spins=main_spins, J=J)
update_spin_1 = isi.update_main_spins_1(ind1=ind1, main_spins=main_spins, param_T=param_T,
energy_net=energy_net)
update_spin_2 = isi.update_main_spins_2(ind2=ind2, main_spins=main_spins, param_T=param_T,
energy_net=energy_net)
mag = isi.get_magnetization(main_spins)
e = isi.get_energy(main_spins)
### END -- OPEN ISING MODEL ###
t_schd = t_schd_in
assert (len(t_schd) == t_slots)
# CREATE LOG FILE FOR RECORDING SIM OUTPUTS/PARAMETERS
log_filename = 'Data_Log_ver_' + str(save_ver)
log_fn = os.path.abspath(os.path.join(dataPath_ver, log_filename))
with open(log_fn, 'w') as f:
## COMPUTE ms BETWEEN SPIKES BASED ON ASTROCYTE TIME
min_stim_period = int(1000 / max_rate)
## COMPUTE NUMBER OF ISING UPDATES PER TEMP BASED ON ASTROCYTE TIME
ising_updates_per_temp = int(ms_per_temp / min_stim_period)
ising_updates_per_ini = int(ini_dur_ms / min_stim_period)
if ising_updates_per_temp * min_stim_period != ms_per_temp:
ms_per_temp = min_stim_period * ising_updates_per_temp
if ising_updates_per_ini * min_stim_period != ini_dur_ms:
ini_dur_ms = min_stim_period * ising_updates_per_ini
f.write('LOG____v_' + str(save_ver) + '\n\n')
f.write('MODEL_______' + str(model) + '\n\n')
f.write('DATA PATH: ' + str(dataPath_ver) + '\n\n\n')
f.write('------------------------------------------------------------\n\n')
f.write('INPUT PARAMETERS:\n\n')
f.write(' ms_per_temp = ' + str(ms_per_temp) + '\n')
f.write(' ini_dur_ini = ' + str(ini_dur_ms) + '\n')
f.write(' t_slots = ' + str(t_slots) + '\n')
f.write(' max_rate' + str(max_rate) + '\n')
f.write(' v_beta_scale_factor = ' + str(v_beta_scale_factor) + '\n')
f.write(' randomize_glut_amount = ' + str(randomize_glut_amount) + '\n')
f.write(' glut_level = ' + str(glut_level) + '\n')
f.write('\n\n')
f.write('RUNTIME PARAMETERS AND STATS:\n\n')
f.write(' Max Rate: ' + str(max_rate) + '\n')
f.write(' min_stim_period: ' + str(min_stim_period) + '\n')
f.write(' ising_updates_per_temp: ' + str(ising_updates_per_temp) + '\n')
f.write(' ising_updates_per_ini: ' + str(ising_updates_per_ini) + '\n\n')
f.write('Temp Schedule: \n')
f.write(str(t_schd) + '\n\n')
assert (ising_updates_per_temp * min_stim_period == ms_per_temp)
assert (ising_updates_per_ini * min_stim_period == ini_dur_ms)
#### INITIALIZE GCH-I MODEL ##### START
astro = astr_astro.gpu_astro(num_astros=self.astro_num,
max_syns_per_astro=shp[0] * shp[1],
input_morph_matrix=np.ones(
(self.astro_num, shp[0] * shp[1]),
dtype=np.float32))
ip3_state, ca_state, h_state, ip3_store, ca_store = astro.initialize_vars()
syn_input_feed, input_morph_feed, v_beta_feed = astro.initialize_ph()
# state update graph
new_ca_var_states = astro.run_ca_state_transition(L1_ip3_state=ip3_state, L1_ca_state=ca_state,
L1_h_state=h_state)
store_ca_var_states = tf.assign(ca_store, new_ca_var_states)
new_h_var_states = astro.run_h_state_transition(L1_ip3_state=ip3_state, L1_ca_state=ca_state,
L1_h_state=h_state)
update_h_var_states = tf.assign(h_state, new_h_var_states)
new_ip3_var_states = astro.run_ip3_state_transition(L1_ip3_state=ip3_state, L1_ca_state=ca_state,
syn_inp=syn_input_feed,
v_beta_var=v_beta_feed,
input_morph=input_morph_feed)
update_ip3_var_states = tf.assign(ip3_state, new_ip3_var_states)
update_ca_var_states = tf.assign(ca_state, ca_store)
rst_ip_state = astro.reset_var(var=ip3_state,new_val_scalar=astro.ip3_init)
rst_ca_state = astro.reset_var(var=ca_state, new_val_scalar=astro.ca_init)
rst_h_state = astro.reset_var(var=h_state, new_val_scalar=astro.h_init)
#### INITIALIZE GCH-I MODEL ##### END
# -SEEDSET- SET RANDOM SEED FOR TF OPS
# tf.set_random_seed(1234)
### INITIALIZE TF SESSION ###
sess = tf.Session()
sess.run(tf.global_variables_initializer())
f.write('------------------------------------------------------\n\n\n')
# INITIALIZE SAVE LISTS
mag_list = []
e_list = []
temp_list = []
ca_list = []
ising_mat_save = np.zeros((t_slots, shp[0], shp[1]), dtype=np.float32)
spin_rate_per_t_save = np.zeros(np.shape(ising_mat_save), dtype=np.float32)
mag_per_temp_list = []
ms_temp = []
ms_ising = []
ms_time_main = []
# INITIALIZING RANDOM SPIN STATE
print('Setting Random Initial Spin Configuration')
out_rescaled = sess.run(spin_dist)
spins_matrix_rescaled1 = np.add(1, np.multiply(2, np.floor(np.clip(out_rescaled, -1, 0))))
spins_matrix_rescaled = np.subtract(np.multiply(2, np.random.randint(0, 2, size=(np.shape(spins_matrix_rescaled1)[0], np.shape(spins_matrix_rescaled1)[1]))), 1)
sess.run(isi.reinitialize_spins(main_spins=main_spins, feed_spins=spins_matrix_rescaled))
### INITIALIZATION LOOP ###
# set temp for loop iteration
sess.run(isi.set_temperature(param_T=param_T, feed_temp=t_schd[0]))
print('Starting Initialization at T = ' + str(t_schd[0]))
f.write('INITIALIZATION STATS:\n\n')
f.write('Initialization Temperature: ' + str(t_schd[0]) + '\n')
### LETTING ISING SYSTEM STABALIZE
for k in range(0,10000):
sess.run(update_spin_1)
sess.run(update_spin_2)
### GETTING INITIALIZATION RATES
mag_ini_ave = []
temp_spin_agg = np.zeros(shp, dtype=np.float32)
for i in range(0, ising_updates_per_ini):
## Ising Spin Updates
sess.run(update_spin_1)
sess.run(update_spin_2)
out_spins = sess.run(main_spins)
temp_spin_agg += np.clip(out_spins, 0, 1)
mag_ini_ave.append(sess.run(mag))
print('Initialization Complete. Stats: ')
# COMPUTING INITIALIZATION RATES USED TO INITIALIZE GCH-I v_beta's
rates_raw = np.divide(temp_spin_agg, ini_dur_ms / 1000)
rates_ini = self.format_rates(rates_raw, 0.2, max_rate)
f.write('Sum Spin Stats: \n')
f.write(' Average: ' + str(np.average(temp_spin_agg)) + '\n')
f.write(' Maximum: ' + str(np.amax(temp_spin_agg)) + '\n')
f.write(' Minimum: ' + str(np.amin(temp_spin_agg)) + '\n\n')
f.write('Spin Rates_Raw Stats: \n')
f.write(' Average: ' + str(np.average(rates_raw)) + '\n')
f.write(' Maximum: ' + str(np.amax(rates_raw)) + '\n')
f.write(' Minimum: ' + str(np.amin(rates_raw)) + '\n\n')
f.write('Spin Rates Stats: \n')
f.write(' Average: ' + str(np.average(rates_ini)) + '\n')
f.write(' Maximum: ' + str(np.amax(rates_ini)) + '\n')
f.write(' Minimum: ' + str(np.amin(rates_ini)) + '\n\n')
### INITIALIZATION OF V_BETA WITH GLUT_LEVEL ###
for i in range(0,len(v_beta_scale_factor)):
f.write('v_beta for SCALING FACTOR: ' + str(v_beta_scale_factor[i]) + ' GLUT_LEVEL: ' + str(glut_level[i]) + '\n\n')
v_beta = self.rates_to_v_beta(rates_ini, scaling=v_beta_scale_factor[i],shp=shp)
print('v_beta Ave, Min, Max: ', np.average(v_beta), np.amin(v_beta), np.amax(v_beta))
f.write('v_beta Stats: \n')
f.write(' Average: ' + str(np.average(v_beta)) + '\n')
f.write(' Maximum: ' + str(np.amax(v_beta)) + '\n')
f.write(' Minimum: ' + str(np.amin(v_beta)) + '\n\n')
if i == 0:
## PREPARE V_BETA FOR INPUT INTO GPU ##
v_beta_f = v_beta.reshape(1, -1)
rates_ini_f = rates_ini.reshape(1, -1)
else:
v_beta_f = np.concatenate([v_beta_f, v_beta.reshape(1, -1)],axis=0)
rates_ini_f = np.concatenate([rates_ini_f, rates_ini.reshape(1, -1)],axis=0)
##### REFORMAT ASTRO INPUTS
v_beta_f_rsc = np.divide(v_beta_f,1000)
zero_input = np.zeros(np.shape(v_beta_f), dtype=np.float32)
morph = np.ones(np.shape(v_beta_f), dtype=np.float32)
glut_level_weights = np.broadcast_to(np.asarray(glut_level).reshape(-1,1),shape=np.shape(v_beta_f))
f.write('-------------------------------------------------\n\n\n')
f.write('SIMULATION OUTPUT AND RESULTS: \n\n')
# T LOOP
for i in range(0, len(t_schd)):
print('Launching T' + str(i) + '/' + str(len(t_schd)) + ': ' + str(t_schd[i]))
f.write('Launching T' + str(i) + '/' + str(len(t_schd)) + ': ' + str(t_schd[i]) + '\n')
# set temp for loop iteration
sess.run(isi.set_temperature(param_T=param_T, feed_temp=t_schd[i]))
# RESETS GCH-I STATE VARIABLES IF T CHANGES
if i > 0:
if np.absolute(t_schd[i] - t_schd[i-1]) > 0:
sess.run(rst_ca_state)
sess.run(rst_ip_state)
sess.run(rst_h_state)
st = time.time()
mag_total_per_temp = 0
temp_spin_agg = np.multiply(temp_spin_agg, 0.0)
# ISING LOOP
for t in range(0, ising_updates_per_temp):
## Ising Spin Updates
sess.run(update_spin_1)
sess.run(update_spin_2)
# GET AND REFORMAT SPINS
out_spins = sess.run(main_spins)
out_spins_fmt = np.clip(out_spins, 0, 1)
out_spins_f = np.broadcast_to(out_spins_fmt.reshape(1, -1), shape=np.shape(glut_level_weights))
out_spins_f2 = np.multiply(glut_level_weights,out_spins_f)
# ASTRO LOOP
for a in range(0, min_stim_period):
# COMPUTE CURRENT ASTROCYTE TIME
cur_ms = (i * ising_updates_per_temp * min_stim_period) + (t * min_stim_period) + a
ms_time_main.append(cur_ms)
if a == 0:
### UPDATE ASTRO WITH ISING INPUT
sess.run(store_ca_var_states)
sess.run(update_h_var_states)
sess.run(update_ip3_var_states, feed_dict={syn_input_feed: out_spins_f2,
v_beta_feed: v_beta_f_rsc,
input_morph_feed: morph})
sess.run(update_ca_var_states)
else:
### UPDATE ASTRO WITHOUT ISING INPUT
sess.run(store_ca_var_states)
sess.run(update_h_var_states)
sess.run(update_ip3_var_states,
feed_dict={syn_input_feed: zero_input,
v_beta_feed: v_beta_f_rsc,
input_morph_feed: morph})
sess.run(update_ca_var_states)
# RECORD OUTPUTS
mag_list.append(sess.run(mag))
mag_total_per_temp += sess.run(mag)
e_list.append(sess.run(e))
temp_list.append(t_schd[i])
ms_ising.append(cur_ms)
temp_spin_agg += out_spins_fmt
ca_list.append(sess.run(ca_state))
mag_per_temp_list.append(np.divide(mag_total_per_temp, ising_updates_per_temp))
# COMPUTE AND RECORD SPIN FLIP RATES
rates = np.divide(temp_spin_agg, ms_per_temp / 1000)
spin_rate_per_t_save[i, :, :] = rates
ms_temp.append(cur_ms)
print('T_' + str(t_schd[i]) + '_Complete_in_' + str(time.time() - st) + '_sec')
f.write('T_' + str(t_schd[i]) + '_Complete_in_' + str(time.time() - st) + '_sec\n')
sess.close()
data_name = 'ICA_Data_ver_' + str(save_ver)
sup.save_non_tf_data(names = ['t_schd','mag_list','temp_list','ms_ising','ca_list','spin_rate_per_t_save','e_list']
, data = [t_schd,mag_list,temp_list,ms_ising,ca_list,spin_rate_per_t_save,e_list]
, filename=data_name
, savePath=dataPath_ver
)
print('File ' + str(data_name) + ' saved.')
def rates_to_v_beta(self,rates,scaling,shp):
'''
:param rates: SPIN FLIP RATES COMPUTED BASED ON ASTROCYTE TIME
:param scaling: MULTIPLICATIVE SCALING FACTOR
:param shp: SHAPE OF LATTICE
:return:
'''
p_hist_mat = np.divide(rates,1000)
v_mat_scaled = np.multiply(np.divide(1,scaling),np.divide(1, p_hist_mat))
return np.clip(np.divide(v_mat_scaled, shp[0] * shp[1]),0.0004,0.13)
def format_rates(self,rates,min_rate_hz,max_rate_hz):
# CLIPS RATE TO min_rate_hz AND max_rate_hz
return np.clip(rates, min_rate_hz, max_rate_hz)
def main(argv):
try:
opts, args = getopt.getopt(argv, "", ["ver_num=", "ising_model_num=", "t1=", "t2="])
except getopt.GetoptError:
print('Incorrect arguments')
sys.exit(2)
for opt, arg in opts:
if opt == '--ver_num':
ver = int(arg)
elif opt == '--ising_model_num':
m = int(arg)
elif opt == '--t1':
t1 = float(arg)
elif opt == '--t2':
t2 = float(arg)
else:
print('Error, exiting')
sys.exit()
####### UNCOMMENT IF USING GPU #######
# GPU_NUM = input('GPU? ')
#
# os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID" # see issue #152
# os.environ["CUDA_VISIBLE_DEVICES"] = str(GPU_NUM)
# print('GPU: ' + str(os.environ["CUDA_VISIBLE_DEVICES"]))
v_beta_scale_factor_list = [5.5]
glut_list = [0.2]
vbsfl = []
gl = []
for i in range(0, len(v_beta_scale_factor_list)):
for j in range(0, len(glut_list)):
vbsfl.append(v_beta_scale_factor_list[i])
gl.append(glut_list[j])
print(vbsfl)
print(gl)
t_list = [t1,t2]
t_slots = 18
t_schd = np.concatenate([t_list[0] * np.ones(int(t_slots / 2)), t_list[1] * np.ones(int(t_slots / 2))],
axis=0) # just two temps
am = astro_pp_model(num_of_astros=len(vbsfl))
am.test_ising_w_patterns_w_astro_v5_5(model=m, save_ver=ver, v_beta_scale_factor=vbsfl,
t_schd_in=t_schd, t_slots=t_slots, randomize_glut_amount=False,
ms_per_temp=15000, max_rate=20, glut_level=gl)
if __name__ == '__main__':
main(sys.argv[1:])