-
Notifications
You must be signed in to change notification settings - Fork 0
/
inputs.py
284 lines (242 loc) · 10.8 KB
/
inputs.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
#!/usr/bin/env python
from config import Configurable, quantity
import brian
import heapq
import numpy as np
import numpy.random as rnd
import quantities as pq
import spykeutils.spike_train_generation as stg
import tables
class InputSignalGenerator(Configurable):
def __init__(self, config, duration=None):
Configurable.__init__(self, config)
self._add_config_value('peak_firing_rate', quantity)
self._add_config_value('background_activity', quantity)
self._add_config_value('sparseness', int)
self._add_config_value('approximate_normalization', float)
self._add_config_value('filter_time_constant', quantity)
self._add_config_value('dt', quantity)
self.current_raw_value = 0.0
self.current_time = 0.0 * brian.second
self.sparsification_start = 1
self.signal = None
self.duration = duration
def gen_filtered_white_noise(self, initial_value, size):
r = rnd.rand(size) - 0.5
filter_value = np.exp(-self.dt / self.filter_time_constant)
signal = np.empty(size)
signal[0] = (1 - filter_value) * r[0] + filter_value * initial_value
for i in xrange(1, signal.size):
signal[i] = (1 - filter_value) * r[i] + filter_value * signal[i - 1]
return signal
def next_interval(self):
if self.duration is None:
duration = 4 * brian.second
self.signal = None
else:
duration = self.duration
if self.signal is None:
self.signal = self.gen_filtered_white_noise(
self.current_raw_value, duration / self.dt)
self.current_raw_value = self.signal[-1]
self.rectify(self.signal)
self.current_time += duration
return self.signal
def rectify(self, signal):
self.remove_bumps(signal)
np.maximum(0, signal, out=signal)
signal *= self.peak_firing_rate * self.dt / \
self.approximate_normalization
signal += self.background_activity * self.dt
def remove_bumps(self, signal):
start = self.sparsification_start
bump_borders = np.diff(np.asarray(signal > 0, dtype=int))
bump_starts = np.nonzero(bump_borders == 1)[0]
bump_ends = np.nonzero(bump_borders == -1)[0]
while bump_starts.size > 0 and bump_ends.size > 0 \
and bump_starts[0] > bump_ends[0]:
if start <= 0:
signal[0:bump_ends[0] + 1] = 0
start += self.sparseness - 1
else:
start = start - 1
bump_ends = bump_ends[1:]
self.sparsification_start = self.sparseness - 1 - \
(bump_starts.size - start) % self.sparseness
for i in xrange(start, bump_starts.size, self.sparseness):
bump_start = bump_starts[i] + 1
if i < bump_ends.size:
bump_end = bump_ends[i] + 1
else:
bump_end = bump_ends.size
self.sparsification_start = 0
signal[bump_start:bump_end] = 0
class PoissonSpikeTimesGenerator(object):
def __init__(self, signal_gen, neuron_indices, refractory_period):
self.signal_gen = signal_gen
self.neuron_indices = neuron_indices
self.refractory_period = refractory_period
self.last_spikes = np.empty(neuron_indices.size)
self.last_spikes.fill(-self.refractory_period)
self.__spikes = []
self.__current_index = -1
def __iter__(self):
return self
def __fill_spikes_for_next_interval(self):
t_starts = np.maximum(
self.signal_gen.current_time,
self.last_spikes + self.refractory_period)
signal = self.signal_gen.next_interval()
t_stop = self.signal_gen.current_time
trains = [self.gen_spike_train(signal, t_starts[i], t_stop).rescale(
pq.s).magnitude for i in xrange(len(self.neuron_indices))]
self.last_spikes = np.asarray(
[st[-1] if st.size > 0 else self.last_spikes[i]
for i, st in enumerate(trains)]) * brian.second
self.__spikes = sorted(
(st * brian.second, i)
for i, train in zip(self.neuron_indices, trains) for st in train)
def gen_spike_train(self, input_signal, t_start, t_stop):
t_start /= brian.second / pq.s
t_stop /= brian.second / pq.s
max_input = input_signal.max()
dt = self.signal_gen.dt
max_rate = max_input / self.signal_gen.dt
input_signal_duration = input_signal.size * dt
input_signal_times = self.signal_gen.current_time - \
input_signal_duration + np.arange(input_signal.size) * dt
def modulation(ts):
return np.interp(
ts.rescale(pq.s).magnitude, input_signal_times,
input_signal) / max_input
return stg.gen_inhomogeneous_poisson(
modulation, max_rate / brian.hertz * pq.Hz,
t_start=t_start, t_stop=t_stop,
refractory=self.refractory_period / brian.second * pq.s)
def next(self):
self.__current_index += 1
if self.__current_index >= len(self.__spikes):
self.__fill_spikes_for_next_interval()
self.__current_index = 0
return self.__spikes[self.__current_index]
class GroupedSpikeTimesGenerator(Configurable):
def __init__(self, config, duration=None):
Configurable.__init__(self, config)
self._add_config_value('num_tunings', int)
self._add_config_value('num_trains', int)
self._add_config_value('fraction_inhibitory', float)
self._add_config_value('refractory_period', quantity)
self.num_trains_per_tuning = self.num_trains // self.num_tunings
self.num_inhibitory = int(self.fraction_inhibitory * self.num_trains)
self.num_excitatory = self.num_trains - self.num_inhibitory
self.num_inhib_per_tuning = self.num_inhibitory / self.num_tunings
self.num_excit_per_tuning = self.num_excitatory / self.num_tunings
self.signal_gens = [
InputSignalGenerator(self._config['raw_signals'], duration)
for i in xrange(self.num_tunings)]
time_gens = [PoissonSpikeTimesGenerator(
gen, self.neuron_indices_of_group(i), self.refractory_period)
for i, gen in enumerate(self.signal_gens)]
self.merged_times = heapq.merge(*time_gens)
def __iter__(self):
return self.merged_times
def excitatory_neuron_indices_of_group(self, n):
start = n * self.num_excit_per_tuning
return np.arange(start, start + self.num_excit_per_tuning)
def inhibitory_neuron_indices_of_group(self, n):
start = n * self.num_inhib_per_tuning + self.num_excitatory
return np.arange(start, start + self.num_inhib_per_tuning)
def neuron_indices_of_group(self, n):
return np.hstack((
self.excitatory_neuron_indices_of_group(n),
self.inhibitory_neuron_indices_of_group(n)))
def get_indexing_scheme(self):
return {
'excitatory': [self.excitatory_neuron_indices_of_group(i)
for i in xrange(self.num_tunings)],
'inhibitory': [self.inhibitory_neuron_indices_of_group(i)
for i in xrange(self.num_tunings)]
}
class StoredSpikeTimesProvider(object):
def __init__(self, data):
self.num_tunings = data.root.input_data.indexing.excitatory._v_nchildren
assert self.num_tunings == \
data.root.input_data.indexing.inhibitory._v_nchildren
self.num_trains = \
sum(len(g) for g in data.root.input_data.indexing.excitatory) + \
sum(len(g) for g in data.root.input_data.indexing.inhibitory)
self.num_trains_per_tuning = self.num_trains // self.num_tunings
self.data = data
self._iter = self.data.root.input_data.spiketimes.iterrows()
def __iter__(self):
return self
def next(self):
row = self._iter.next()
return row['neuron_index'], row['spiketime'] * brian.second
def get_indexing_scheme(self):
return {
'excitatory': [
self.data.root.input_data.indexing.excitatory._f_getChild(
'group%i' % i) for i in xrange(self.num_tunings)],
'inhibitory': [
self.data.root.input_data.indexing.inhibitory._f_getChild(
'group%i' % i) for i in xrange(self.num_tunings)]
}
def swap_tuple_values(tuples):
for x, y in tuples:
yield y, x
class SpiketimesTable(tables.IsDescription):
neuron_index = tables.UIntCol()
spiketime = tables.Float32Col()
if __name__ == '__main__':
import argparse
import json
parser = argparse.ArgumentParser(
description="Produce input spike times for the Vogels et al. 2011 " +
"model.")
parser.add_argument(
'-c', '--config', type=str, nargs=1, required=True,
help="Path to the configuration file.")
parser.add_argument(
'-t', '--time', type=float, nargs=1, required=True,
help="Time point in seconds up to which to generate spikes.")
parser.add_argument(
'output', nargs=1, type=str,
help="Path ot the HDF5 output file.")
parser.add_argument(
'label', nargs='?', type=str,
help="Label for the simulation. Not used!")
args = parser.parse_args()
with open(args.config[0], 'r') as f:
config = json.load(f)
with tables.openFile(args.output[0], 'w') as out_file:
out_file.setNodeAttr('/', 'config', config)
input_group = out_file.createGroup('/', 'input_data', "Input data")
spike_table = out_file.createTable(
input_group, 'spiketimes', SpiketimesTable,
"Table of spike times of Poisson spike train and the index of the" +
" neuron producing the spike.")
spike_table.attrs.spiketime_unit = 'second'
generator = GroupedSpikeTimesGenerator(config)
geniter = iter(generator)
t = 0 * brian.second
i = 0
while t < args.time[0] * brian.second:
t, idx = geniter.next()
spike = spike_table.row
spike['neuron_index'] = idx
spike['spiketime'] = t
spike.append()
i += 1
if i % 1000:
spike_table.flush()
out_file.flush()
indexing_group = out_file.createGroup(
input_group, 'indexing', "Indices of excitatory and inhibitory " +
"neurons for the different tuning groups.")
for key, groups in generator.get_indexing_scheme().iteritems():
for i, group in enumerate(groups):
path = '%s/%s' % (indexing_group._v_pathname, key)
out_file.createArray(
path, 'group%i' % i, group, createparents=True)
out_file.flush()