-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathcurrent_input.py
230 lines (159 loc) · 8.06 KB
/
current_input.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
""" A script to use our spike trainer to learn a current -> spiking relationship """
from spike_trainer_1 import spike_trainer
from neuron import IAF
#from brian_neurons import poisson_neuron
import numpy as np
from matplotlib import pyplot as plt
import random
import ipdb
def poisson_input(nb_epoch=7,num_steps=20000):
hz = 60
(pn,sn,v) = poisson_neuron(hz,num_steps)
# go from spike times to zeros
in_spikes = np.zeros((num_steps,1))
out_spikes = np.zeros((num_steps,1))
for s in sn:
out_spikes[int((s*1000)) ] = 1
for s in pn:
in_spikes[int((s*1000)) ] = 1
st = spike_trainer(in_spikes,out_spikes,nb_epoch=nb_epoch)
st.fit_model()
"""
num_steps = 500
(pn,sn,v) = poisson_neuron(hz,num_steps)
in_spikes = np.zeros((num_steps,1))
out_spikes = np.zeros((num_steps,1))
for s in sn:
out_spikes[int((s*1000)) ] = 1
for s in pn:
in_spikes[int((s*1000)) ] = 1
"""
predicted_spikes = np.array(st.compute(in_spikes[1000:2000]))
print "Ratio = " + str( np.sum(predicted_spikes) / np.sum(out_spikes[1100:2000]))
plt.plot(out_spikes[1100:2000]); plt.plot(-predicted_spikes);plt.show()
plt.plot(out_spikes[1100:2000]); plt.plot(-in_spikes[1100:2000]); plt.show()
def exp_constant_input(nb_epoch=7,num_steps=10000,test_steps=1000,exp_name='exp_constant_input',load_model=True,load_weights=True):
#Training
current = np.ones((num_steps,1)) * 0.1
threshold = 1
(spikes,trace,string) = IAF(current,threshold)
st = spike_trainer(trace,spikes,nb_epoch=nb_epoch,exp_name=exp_name,load_model=load_model,load_weights=load_weights)
st.fit_model()
test_levels = [0.1]
results = [{} for l in test_levels]
for i, level in enumerate(test_levels):
# Prediction
print "Predict at " + str(level)
current = np.ones((test_steps,1)) * level
(spikes,trace,string) = IAF(current,threshold)
predicted_spikes = st.compute(trace)
results[i]['trace'] = trace
results[i]['actual'] = spikes
results[i]['predicted'] = predicted_spikes
print "Ratio for " +str(level) + " = " + str( np.sum(predicted_spikes) / np.sum(spikes[100:]))
plt.plot(spikes[100:]); plt.plot(predicted_spikes); plt.show()
def exp_variable_input(nb_epoch=7,num_steps=10000,test_steps=1000,exp_name='exp_variable_input',load_model=True,load_weights=True):
#Training
current = np.ones((num_steps,1)) * 0.1
current[2500:5000] = 0.2
current[5000:7500] = 0.3
current[7500:] = 0.4
threshold = 1
(spikes,trace,string) = IAF(current,threshold)
st = spike_trainer(trace,spikes,nb_epoch=nb_epoch,exp_name = exp_name,load_model=load_model,load_weights = load_weights)
st.fit_model()
test_levels = [0.1,0.2,0.3,0.4]
results = [{} for l in test_levels]
for i, level in enumerate(test_levels):
# Prediction
print "Predict at " + str(level)
current = np.ones((test_steps,1)) * level
(spikes,trace,string) = IAF(current,threshold)
predicted_spikes = st.compute(trace)
results[i]['trace'] = trace
results[i]['actual'] = spikes
results[i]['predicted'] = predicted_spikes
print "Ratio for " +str(level) + " = " + str( np.sum(predicted_spikes) / np.sum(spikes[100:]))
plt.plot(spikes[100:]); plt.plot(predicted_spikes); plt.show()
current = np.ones((num_steps,1)) * 0.1
current[250:] = 0.3
(spikes,trace,string) = IAF(current,threshold)
predicted_spikes = st.compute(trace)
print "Ratio for " +str(level) + " = " + str( np.sum(predicted_spikes) / np.sum(spikes[100:]))
plt.plot(spikes[100:]); plt.plot(predicted_spikes); plt.show()
def exp_paired_spikes(nb_epoch=7,num_steps=10000,test_steps=1000,exp_name='exp_paired_spikes_input',load_model=True,load_weights=True):
#Training
def gen_input(num_steps):
in_spikes = np.zeros((num_steps,1))
out_spikes = np.zeros((num_steps,1))
i = 10
while i < num_steps:
i+i+1
if random.random() < 0.05:
in_spikes[i-1] = 1
offset = 2+int(random.random()*10)
in_spikes[i-offset] = 1
out_spikes[i] = 1
i = i +20 + int(random.random()*50)
return (in_spikes,out_spikes)
in_spikes,out_spikes = gen_input(num_steps)
in_spikes_test,out_spikes_test = gen_input(test_steps)
st = spike_trainer(in_spikes,out_spikes,exp_name=exp_name,nb_epoch=nb_epoch,load_model=load_model,load_weights=load_weights)
st.fit_model()
predicted_spikes = st.compute(in_spikes_test)
print "Ratio = " + str( np.sum(predicted_spikes) / np.sum(out_spikes_test[100:]))
#plt.plot(out_spikes_test[100:]); plt.plot(-in_spikes_test[100:]); plt.show()
plt.plot(out_spikes_test[100:]); plt.plot(-np.array(predicted_spikes)); plt.show()
def exp_spiking_IAF_input(nb_epoch,num_steps=10000,test_steps=1000,exp_name='exp_spiking_IAF_input',load_model=True,load_weights=True):
in_spikes = np.zeros((num_steps,1))
out_spikes = np.zeros((num_steps,1))
def gen_spikes(num_steps):
in_spikes = np.zeros((num_steps,1))
out_spikes = np.zeros((num_steps,1))
s = 0
s_t = 2
isi = 0
for i in np.arange(num_steps-1):
if random.random() <0.1:
in_spikes[i] = 1
s = s+1
if s == s_t:
out_spikes[i+1] = 1
s = 0
in_spikes_filter = np.zeros_like(in_spikes)
def filter(a):
f = np.zeros_like(a)
for i in np.arange(1,num_steps):
f[i] = 0.80*f[i-1] + a[i]
return f/np.max(f)
return (filter(in_spikes), filter(out_spikes))
in_spikes,out_spikes = gen_spikes(num_steps)
in_spikes_test,out_spikes_test = gen_spikes(test_steps)
st = spike_trainer(in_spikes,out_spikes,exp_name=exp_name,nb_epoch = nb_epoch,load_weights=load_weights,load_model=load_model)
st.fit_model()
predicted_spikes = st.compute(in_spikes_test)
print "reproduced " + str(sum(predicted_spikes)) + " of " + str(sum(out_spikes_test[100:]))
plt.plot(out_spikes_test[100:]); plt.plot(predicted_spikes); plt.show()
def exp_spiking_eIAF_input(nb_epoch,num_steps=10000,test_steps=1000,exp_name='exp_spiking_eIAF_input',load_model=True,load_weights=True):
""" using a basic sypnape integrate and fire model """
from neuron import eIAF
in_spikes = 1.0* (np.random.rand(num_steps,1) < 0.1)
in_spikes_test = 1.0* (np.random.rand(test_steps,1) < 0.1)
tau = 0.8
threshold = 1
out_spikes,out_filter,in_filter = eIAF(in_spikes,1,tau)
out_spikes_test,out_filter_test,in_filter_test = eIAF(in_spikes_test,1,tau)
st = spike_trainer(in_filter,out_filter,exp_name=exp_name,nb_epoch = nb_epoch,load_weights=load_weights,load_model=load_model)
st.fit_model()
predicted_spikes = st.compute(in_filter_test)
print "reproduced " + str(sum(predicted_spikes)) + " of " + str(sum(out_filter_test[100:]))
plt.plot(out_spikes_test[100:]); plt.plot(predicted_spikes); plt.show()
plt.plot(out_filter_test[100:]); plt.plot(predicted_spikes); plt.show()
ipdb.set_trace()
if __name__ == "__main__":
#exp_constant_input(nb_epoch = 10,num_steps = 10000,test_steps=500, exp_name = 'exp_constant_input',load_model=False,load_weights =False)
#exp_variable_input(nb_epoch = 3,num_steps = 10000, exp_name = 'exp_variable_input',load_model = True,load_weights=True)
#poisson_input(nb_epoch = 3,num_steps = 10000, exp_name = 'exp_poisson_input')
#exp_paired_spikes(nb_epoch = 7,num_steps = 10000, exp_name = 'exp_paired_spikes_input',load_model=True,load_weights=True)
#exp_spiking_IAF_input(nb_epoch = 5,num_steps = 10000, exp_name = 'exp_spiking_IAF_input_mse_relu_full',load_model=True,load_weights=True)
exp_spiking_eIAF_input(nb_epoch = 1,num_steps = 5000,test_steps=2000, exp_name = 'exp_spiking_eIAF_inputi_filter', load_model=True, load_weights=True)