-
Notifications
You must be signed in to change notification settings - Fork 139
/
Diehl&Cook_spiking_MNIST.py
537 lines (469 loc) · 21.6 KB
/
Diehl&Cook_spiking_MNIST.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
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
'''
Created on 15.12.2014
@author: Peter U. Diehl
'''
import numpy as np
import matplotlib.cm as cmap
import time
import os.path
import scipy
import cPickle as pickle
import brian_no_units #import it to deactivate unit checking --> This should NOT be done for testing/debugging
import brian as b
from struct import unpack
from brian import *
# specify the location of the MNIST data
MNIST_data_path = ''
#------------------------------------------------------------------------------
# functions
#------------------------------------------------------------------------------
def get_labeled_data(picklename, bTrain = True):
"""Read input-vector (image) and target class (label, 0-9) and return
it as list of tuples.
"""
if os.path.isfile('%s.pickle' % picklename):
data = pickle.load(open('%s.pickle' % picklename))
else:
# Open the images with gzip in read binary mode
if bTrain:
images = open(MNIST_data_path + 'train-images.idx3-ubyte','rb')
labels = open(MNIST_data_path + 'train-labels.idx1-ubyte','rb')
else:
images = open(MNIST_data_path + 't10k-images.idx3-ubyte','rb')
labels = open(MNIST_data_path + 't10k-labels.idx1-ubyte','rb')
# Get metadata for images
images.read(4) # skip the magic_number
number_of_images = unpack('>I', images.read(4))[0]
rows = unpack('>I', images.read(4))[0]
cols = unpack('>I', images.read(4))[0]
# Get metadata for labels
labels.read(4) # skip the magic_number
N = unpack('>I', labels.read(4))[0]
if number_of_images != N:
raise Exception('number of labels did not match the number of images')
# Get the data
x = np.zeros((N, rows, cols), dtype=np.uint8) # Initialize numpy array
y = np.zeros((N, 1), dtype=np.uint8) # Initialize numpy array
for i in xrange(N):
if i % 1000 == 0:
print("i: %i" % i)
x[i] = [[unpack('>B', images.read(1))[0] for unused_col in xrange(cols)] for unused_row in xrange(rows) ]
y[i] = unpack('>B', labels.read(1))[0]
data = {'x': x, 'y': y, 'rows': rows, 'cols': cols}
pickle.dump(data, open("%s.pickle" % picklename, "wb"))
return data
def get_matrix_from_file(fileName):
offset = len(ending) + 4
if fileName[-4-offset] == 'X':
n_src = n_input
else:
if fileName[-3-offset]=='e':
n_src = n_e
else:
n_src = n_i
if fileName[-1-offset]=='e':
n_tgt = n_e
else:
n_tgt = n_i
readout = np.load(fileName)
print readout.shape, fileName
value_arr = np.zeros((n_src, n_tgt))
if not readout.shape == (0,):
value_arr[np.int32(readout[:,0]), np.int32(readout[:,1])] = readout[:,2]
return value_arr
def save_connections(ending = ''):
print 'save connections'
for connName in save_conns:
connMatrix = connections[connName][:]
# connListSparse = ([(i,j[0],j[1]) for i in xrange(connMatrix.shape[0]) for j in zip(connMatrix.rowj[i],connMatrix.rowdata[i])])
connListSparse = ([(i,j,connMatrix[i,j]) for i in xrange(connMatrix.shape[0]) for j in xrange(connMatrix.shape[1]) ])
np.save(data_path + 'weights/' + connName + ending, connListSparse)
def save_theta(ending = ''):
print 'save theta'
for pop_name in population_names:
np.save(data_path + 'weights/theta_' + pop_name + ending, neuron_groups[pop_name + 'e'].theta)
def normalize_weights():
for connName in connections:
if connName[1] == 'e' and connName[3] == 'e':
connection = connections[connName][:]
temp_conn = np.copy(connection)
colSums = np.sum(temp_conn, axis = 0)
colFactors = weight['ee_input']/colSums
for j in xrange(n_e):#
connection[:,j] *= colFactors[j]
def get_2d_input_weights():
name = 'XeAe'
weight_matrix = np.zeros((n_input, n_e))
n_e_sqrt = int(np.sqrt(n_e))
n_in_sqrt = int(np.sqrt(n_input))
num_values_col = n_e_sqrt*n_in_sqrt
num_values_row = num_values_col
rearranged_weights = np.zeros((num_values_col, num_values_row))
connMatrix = connections[name][:]
weight_matrix = np.copy(connMatrix)
for i in xrange(n_e_sqrt):
for j in xrange(n_e_sqrt):
rearranged_weights[i*n_in_sqrt : (i+1)*n_in_sqrt, j*n_in_sqrt : (j+1)*n_in_sqrt] = \
weight_matrix[:, i + j*n_e_sqrt].reshape((n_in_sqrt, n_in_sqrt))
return rearranged_weights
def plot_2d_input_weights():
name = 'XeAe'
weights = get_2d_input_weights()
fig = b.figure(fig_num, figsize = (18, 18))
im2 = b.imshow(weights, interpolation = "nearest", vmin = 0, vmax = wmax_ee, cmap = cmap.get_cmap('hot_r'))
b.colorbar(im2)
b.title('weights of connection' + name)
fig.canvas.draw()
return im2, fig
def update_2d_input_weights(im, fig):
weights = get_2d_input_weights()
im.set_array(weights)
fig.canvas.draw()
return im
def get_current_performance(performance, current_example_num):
current_evaluation = int(current_example_num/update_interval)
start_num = current_example_num - update_interval
end_num = current_example_num
difference = outputNumbers[start_num:end_num, 0] - input_numbers[start_num:end_num]
correct = len(np.where(difference == 0)[0])
performance[current_evaluation] = correct / float(update_interval) * 100
return performance
def plot_performance(fig_num):
num_evaluations = int(num_examples/update_interval)
time_steps = range(0, num_evaluations)
performance = np.zeros(num_evaluations)
fig = b.figure(fig_num, figsize = (5, 5))
fig_num += 1
ax = fig.add_subplot(111)
im2, = ax.plot(time_steps, performance) #my_cmap
b.ylim(ymax = 100)
b.title('Classification performance')
fig.canvas.draw()
return im2, performance, fig_num, fig
def update_performance_plot(im, performance, current_example_num, fig):
performance = get_current_performance(performance, current_example_num)
im.set_ydata(performance)
fig.canvas.draw()
return im, performance
def get_recognized_number_ranking(assignments, spike_rates):
summed_rates = [0] * 10
num_assignments = [0] * 10
for i in xrange(10):
num_assignments[i] = len(np.where(assignments == i)[0])
if num_assignments[i] > 0:
summed_rates[i] = np.sum(spike_rates[assignments == i]) / num_assignments[i]
return np.argsort(summed_rates)[::-1]
def get_new_assignments(result_monitor, input_numbers):
assignments = np.zeros(n_e)
input_nums = np.asarray(input_numbers)
maximum_rate = [0] * n_e
for j in xrange(10):
num_assignments = len(np.where(input_nums == j)[0])
if num_assignments > 0:
rate = np.sum(result_monitor[input_nums == j], axis = 0) / num_assignments
for i in xrange(n_e):
if rate[i] > maximum_rate[i]:
maximum_rate[i] = rate[i]
assignments[i] = j
return assignments
#------------------------------------------------------------------------------
# load MNIST
#------------------------------------------------------------------------------
start = time.time()
training = get_labeled_data(MNIST_data_path + 'training')
end = time.time()
print 'time needed to load training set:', end - start
start = time.time()
testing = get_labeled_data(MNIST_data_path + 'testing', bTrain = False)
end = time.time()
print 'time needed to load test set:', end - start
#------------------------------------------------------------------------------
# set parameters and equations
#------------------------------------------------------------------------------
test_mode = True
b.set_global_preferences(
defaultclock = b.Clock(dt=0.5*b.ms), # The default clock to use if none is provided or defined in any enclosing scope.
useweave = True, # Defines whether or not functions should use inlined compiled C code where defined.
gcc_options = ['-ffast-math -march=native'], # Defines the compiler switches passed to the gcc compiler.
#For gcc versions 4.2+ we recommend using -march=native. By default, the -ffast-math optimizations are turned on
usecodegen = True, # Whether or not to use experimental code generation support.
usecodegenweave = True, # Whether or not to use C with experimental code generation support.
usecodegenstateupdate = True, # Whether or not to use experimental code generation support on state updaters.
usecodegenthreshold = False, # Whether or not to use experimental code generation support on thresholds.
usenewpropagate = True, # Whether or not to use experimental new C propagation functions.
usecstdp = True, # Whether or not to use experimental new C STDP.
)
np.random.seed(0)
data_path = './'
if test_mode:
weight_path = data_path + 'weights/'
num_examples = 10000 * 1
use_testing_set = True
do_plot_performance = False
record_spikes = True
ee_STDP_on = False
update_interval = num_examples
else:
weight_path = data_path + 'random/'
num_examples = 60000 * 3
use_testing_set = False
do_plot_performance = True
if num_examples <= 60000:
record_spikes = True
else:
record_spikes = True
ee_STDP_on = True
ending = ''
n_input = 784
n_e = 400
n_i = n_e
single_example_time = 0.35 * b.second #
resting_time = 0.15 * b.second
runtime = num_examples * (single_example_time + resting_time)
if num_examples <= 10000:
update_interval = num_examples
weight_update_interval = 20
else:
update_interval = 10000
weight_update_interval = 100
if num_examples <= 60000:
save_connections_interval = 10000
else:
save_connections_interval = 10000
update_interval = 10000
v_rest_e = -65. * b.mV
v_rest_i = -60. * b.mV
v_reset_e = -65. * b.mV
v_reset_i = -45. * b.mV
v_thresh_e = -52. * b.mV
v_thresh_i = -40. * b.mV
refrac_e = 5. * b.ms
refrac_i = 2. * b.ms
conn_structure = 'dense'
weight = {}
delay = {}
input_population_names = ['X']
population_names = ['A']
input_connection_names = ['XA']
save_conns = ['XeAe']
input_conn_names = ['ee_input']
recurrent_conn_names = ['ei', 'ie']
weight['ee_input'] = 78.
delay['ee_input'] = (0*b.ms,10*b.ms)
delay['ei_input'] = (0*b.ms,5*b.ms)
input_intensity = 2.
start_input_intensity = input_intensity
tc_pre_ee = 20*b.ms
tc_post_1_ee = 20*b.ms
tc_post_2_ee = 40*b.ms
nu_ee_pre = 0.0001 # learning rate
nu_ee_post = 0.01 # learning rate
wmax_ee = 1.0
exp_ee_pre = 0.2
exp_ee_post = exp_ee_pre
STDP_offset = 0.4
if test_mode:
scr_e = 'v = v_reset_e; timer = 0*ms'
else:
tc_theta = 1e7 * b.ms
theta_plus_e = 0.05 * b.mV
scr_e = 'v = v_reset_e; theta += theta_plus_e; timer = 0*ms'
offset = 20.0*b.mV
v_thresh_e = '(v>(theta - offset + ' + str(v_thresh_e) + ')) * (timer>refrac_e)'
neuron_eqs_e = '''
dv/dt = ((v_rest_e - v) + (I_synE+I_synI) / nS) / (100*ms) : volt
I_synE = ge * nS * -v : amp
I_synI = gi * nS * (-100.*mV-v) : amp
dge/dt = -ge/(1.0*ms) : 1
dgi/dt = -gi/(2.0*ms) : 1
'''
if test_mode:
neuron_eqs_e += '\n theta :volt'
else:
neuron_eqs_e += '\n dtheta/dt = -theta / (tc_theta) : volt'
neuron_eqs_e += '\n dtimer/dt = 100.0 : ms'
neuron_eqs_i = '''
dv/dt = ((v_rest_i - v) + (I_synE+I_synI) / nS) / (10*ms) : volt
I_synE = ge * nS * -v : amp
I_synI = gi * nS * (-85.*mV-v) : amp
dge/dt = -ge/(1.0*ms) : 1
dgi/dt = -gi/(2.0*ms) : 1
'''
eqs_stdp_ee = '''
post2before : 1.0
dpre/dt = -pre/(tc_pre_ee) : 1.0
dpost1/dt = -post1/(tc_post_1_ee) : 1.0
dpost2/dt = -post2/(tc_post_2_ee) : 1.0
'''
eqs_stdp_pre_ee = 'pre = 1.; w -= nu_ee_pre * post1'
eqs_stdp_post_ee = 'post2before = post2; w += nu_ee_post * pre * post2before; post1 = 1.; post2 = 1.'
b.ion()
fig_num = 1
neuron_groups = {}
input_groups = {}
connections = {}
stdp_methods = {}
rate_monitors = {}
spike_monitors = {}
spike_counters = {}
result_monitor = np.zeros((update_interval,n_e))
neuron_groups['e'] = b.NeuronGroup(n_e*len(population_names), neuron_eqs_e, threshold= v_thresh_e, refractory= refrac_e, reset= scr_e,
compile = True, freeze = True)
neuron_groups['i'] = b.NeuronGroup(n_i*len(population_names), neuron_eqs_i, threshold= v_thresh_i, refractory= refrac_i, reset= v_reset_i,
compile = True, freeze = True)
#------------------------------------------------------------------------------
# create network population and recurrent connections
#------------------------------------------------------------------------------
for name in population_names:
print 'create neuron group', name
neuron_groups[name+'e'] = neuron_groups['e'].subgroup(n_e)
neuron_groups[name+'i'] = neuron_groups['i'].subgroup(n_i)
neuron_groups[name+'e'].v = v_rest_e - 40. * b.mV
neuron_groups[name+'i'].v = v_rest_i - 40. * b.mV
if test_mode or weight_path[-8:] == 'weights/':
neuron_groups['e'].theta = np.load(weight_path + 'theta_' + name + ending + '.npy')
else:
neuron_groups['e'].theta = np.ones((n_e)) * 20.0*b.mV
print 'create recurrent connections'
for conn_type in recurrent_conn_names:
connName = name+conn_type[0]+name+conn_type[1]
weightMatrix = get_matrix_from_file(weight_path + '../random/' + connName + ending + '.npy')
connections[connName] = b.Connection(neuron_groups[connName[0:2]], neuron_groups[connName[2:4]], structure= conn_structure,
state = 'g'+conn_type[0])
connections[connName].connect(neuron_groups[connName[0:2]], neuron_groups[connName[2:4]], weightMatrix)
if ee_STDP_on:
if 'ee' in recurrent_conn_names:
stdp_methods[name+'e'+name+'e'] = b.STDP(connections[name+'e'+name+'e'], eqs=eqs_stdp_ee, pre = eqs_stdp_pre_ee,
post = eqs_stdp_post_ee, wmin=0., wmax= wmax_ee)
print 'create monitors for', name
rate_monitors[name+'e'] = b.PopulationRateMonitor(neuron_groups[name+'e'], bin = (single_example_time+resting_time)/b.second)
rate_monitors[name+'i'] = b.PopulationRateMonitor(neuron_groups[name+'i'], bin = (single_example_time+resting_time)/b.second)
spike_counters[name+'e'] = b.SpikeCounter(neuron_groups[name+'e'])
if record_spikes:
spike_monitors[name+'e'] = b.SpikeMonitor(neuron_groups[name+'e'])
spike_monitors[name+'i'] = b.SpikeMonitor(neuron_groups[name+'i'])
if record_spikes:
b.figure(fig_num)
fig_num += 1
b.ion()
b.subplot(211)
b.raster_plot(spike_monitors['Ae'], refresh=1000*b.ms, showlast=1000*b.ms)
b.subplot(212)
b.raster_plot(spike_monitors['Ai'], refresh=1000*b.ms, showlast=1000*b.ms)
#------------------------------------------------------------------------------
# create input population and connections from input populations
#------------------------------------------------------------------------------
pop_values = [0,0,0]
for i,name in enumerate(input_population_names):
input_groups[name+'e'] = b.PoissonGroup(n_input, 0)
rate_monitors[name+'e'] = b.PopulationRateMonitor(input_groups[name+'e'], bin = (single_example_time+resting_time)/b.second)
for name in input_connection_names:
print 'create connections between', name[0], 'and', name[1]
for connType in input_conn_names:
connName = name[0] + connType[0] + name[1] + connType[1]
weightMatrix = get_matrix_from_file(weight_path + connName + ending + '.npy')
connections[connName] = b.Connection(input_groups[connName[0:2]], neuron_groups[connName[2:4]], structure= conn_structure,
state = 'g'+connType[0], delay=True, max_delay=delay[connType][1])
connections[connName].connect(input_groups[connName[0:2]], neuron_groups[connName[2:4]], weightMatrix, delay=delay[connType])
if ee_STDP_on:
print 'create STDP for connection', name[0]+'e'+name[1]+'e'
stdp_methods[name[0]+'e'+name[1]+'e'] = b.STDP(connections[name[0]+'e'+name[1]+'e'], eqs=eqs_stdp_ee, pre = eqs_stdp_pre_ee,
post = eqs_stdp_post_ee, wmin=0., wmax= wmax_ee)
#------------------------------------------------------------------------------
# run the simulation and set inputs
#------------------------------------------------------------------------------
previous_spike_count = np.zeros(n_e)
assignments = np.zeros(n_e)
input_numbers = [0] * num_examples
outputNumbers = np.zeros((num_examples, 10))
if not test_mode:
input_weight_monitor, fig_weights = plot_2d_input_weights()
fig_num += 1
if do_plot_performance:
performance_monitor, performance, fig_num, fig_performance = plot_performance(fig_num)
for i,name in enumerate(input_population_names):
input_groups[name+'e'].rate = 0
b.run(0)
j = 0
while j < (int(num_examples)):
if test_mode:
if use_testing_set:
rates = testing['x'][j%10000,:,:].reshape((n_input)) / 8. * input_intensity
else:
rates = training['x'][j%60000,:,:].reshape((n_input)) / 8. * input_intensity
else:
normalize_weights()
rates = training['x'][j%60000,:,:].reshape((n_input)) / 8. * input_intensity
input_groups['Xe'].rate = rates
# print 'run number:', j+1, 'of', int(num_examples)
b.run(single_example_time, report='text')
if j % update_interval == 0 and j > 0:
assignments = get_new_assignments(result_monitor[:], input_numbers[j-update_interval : j])
if j % weight_update_interval == 0 and not test_mode:
update_2d_input_weights(input_weight_monitor, fig_weights)
if j % save_connections_interval == 0 and j > 0 and not test_mode:
save_connections(str(j))
save_theta(str(j))
current_spike_count = np.asarray(spike_counters['Ae'].count[:]) - previous_spike_count
previous_spike_count = np.copy(spike_counters['Ae'].count[:])
if np.sum(current_spike_count) < 5:
input_intensity += 1
for i,name in enumerate(input_population_names):
input_groups[name+'e'].rate = 0
b.run(resting_time)
else:
result_monitor[j%update_interval,:] = current_spike_count
if test_mode and use_testing_set:
input_numbers[j] = testing['y'][j%10000][0]
else:
input_numbers[j] = training['y'][j%60000][0]
outputNumbers[j,:] = get_recognized_number_ranking(assignments, result_monitor[j%update_interval,:])
if j % 100 == 0 and j > 0:
print 'runs done:', j, 'of', int(num_examples)
if j % update_interval == 0 and j > 0:
if do_plot_performance:
unused, performance = update_performance_plot(performance_monitor, performance, j, fig_performance)
print 'Classification performance', performance[:(j/float(update_interval))+1]
for i,name in enumerate(input_population_names):
input_groups[name+'e'].rate = 0
b.run(resting_time)
input_intensity = start_input_intensity
j += 1
#------------------------------------------------------------------------------
# save results
#------------------------------------------------------------------------------
print 'save results'
if not test_mode:
save_theta()
if not test_mode:
save_connections()
else:
np.save(data_path + 'activity/resultPopVecs' + str(num_examples), result_monitor)
np.save(data_path + 'activity/inputNumbers' + str(num_examples), input_numbers)
#------------------------------------------------------------------------------
# plot results
#------------------------------------------------------------------------------
if rate_monitors:
b.figure(fig_num)
fig_num += 1
for i, name in enumerate(rate_monitors):
b.subplot(len(rate_monitors), 1, i)
b.plot(rate_monitors[name].times/b.second, rate_monitors[name].rate, '.')
b.title('Rates of population ' + name)
if spike_monitors:
b.figure(fig_num)
fig_num += 1
for i, name in enumerate(spike_monitors):
b.subplot(len(spike_monitors), 1, i)
b.raster_plot(spike_monitors[name])
b.title('Spikes of population ' + name)
if spike_counters:
b.figure(fig_num)
fig_num += 1
for i, name in enumerate(spike_counters):
b.subplot(len(spike_counters), 1, i)
b.plot(spike_counters['Ae'].count[:])
b.title('Spike count of population ' + name)
plot_2d_input_weights()
b.ioff()
b.show()