-
Notifications
You must be signed in to change notification settings - Fork 1
/
optimizer.py
363 lines (308 loc) · 15.3 KB
/
optimizer.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
import numpy as np
from bluepyopt.deapext.optimisations import IBEADEAPOptimisation
from bluepyopt.ephys.objectives import WeightedSumObjective, SingletonObjective
from bluepyopt.parameters import Parameter
from bluepyopt.evaluators import Evaluator
from neat import NeuronSimTree
import pickle
from channels import channelcollection
import utils, data
## optimization ################################################################
# trial run parameter values
MAX_ITER = 10
N_OFFSPRING = 2
# full optimization parameter values
# MAX_ITER = 100
# N_OFFSPRING = 100
################################################################################
## model evaluator for optimization ############################################
class ModelEvaluator(Evaluator):
def __init__(self, sim_tree, v_dat,
loc_soma, loc_dend,
channel_names=['L'],
mode='fit'):
"""
if mode is fit, check bounds, if mode is evaluate, dont check bounds
"""
self.sim_tree = sim_tree
self.channel_names = channel_names
# injection sites
self.loc_soma, self.loc_dend = loc_soma, loc_dend
# params
self.dt, self.dur, self.t_max = data.DT, data.DUR, data.T_MAX
self.t0, self.t1, self.t2, self.t3 = data.T0, data.T1, data.T2, data.T3
self.a0, self.a1, self.a2, self.a3 = data.A0, data.A1, data.A2, data.A3
self.channel_names = channel_names
# constant params
self.r_a = 113./1e6
self.g_max = {'L': 20.*1e2, 'L_c': 20.*1e2,
'K_ir': 40.*1e2, 'K_m': 40.*1e2, 'K_m35': 40.*1e2,
'h_u': 60.*1e2, 'h_HAY': 60.*1e2,
'Na_p': 100.*1e2, 'NaP': 100.*1e2}
# define fit parameters
self._defineFitObjects(v_dat)
# don't check bounds
if mode == 'evaluate':
for p in self.params:
p.bounds = None
def _defineFitObjects(self, v_dat):
# fitness evaluator
features = [utils.VeqFeature(v_dat), utils.VStepFeature(v_dat), utils.TraceFeature(v_dat)]
self.objectives = [SingletonObjective('V_eq', features[0]),
SingletonObjective('V_step', features[1]),
SingletonObjective('V_trace', features[2])]
# parameters
self.params = [Parameter('d_c_m', value=0.001, bounds=[0., 0.01]),
Parameter('c_m_0', value=1., bounds=[0.50, 1.50])]
for c_name in self.channel_names:
if c_name == 'K_m' or c_name == 'K_m35' or c_name == 'K_ir':
params = [Parameter('d_'+c_name, value=-1./200., bounds=[-0.1, 0.]),
Parameter('g0_'+c_name, value=40.*1e2, bounds=[0., 10000.]),
Parameter('e_r_'+c_name, value=-85., bounds=[-95.,-80.])]
elif c_name == 'h_HAY' or c_name == 'h_u':
params = [Parameter('d_'+c_name, value=1./200., bounds=[0.0, 0.1]),
Parameter('g0_'+c_name, value=0.0099*1e2, bounds=[0., 10000.]),
Parameter('e_r_'+c_name, value=-40., bounds=[-50.,-30.])]
elif c_name == 'Na_p' or c_name == 'NaP':
params = [Parameter('d_'+c_name, value=0., bounds=[-0.0001, 0.0001]),
Parameter('g0_'+c_name, value=0.01*1e2, bounds=[0.,10000.]),
Parameter('e_r_'+c_name, value=50., bounds=[40.,60.])]
# elif c_name == 'L':
# params = [Parameter('d_L', value=1./200., bounds=[0.0, 0.05]),
# Parameter('g0_L', value=0.40*1e2, bounds=[0., 300.]),
# Parameter('e_0_L', value=-90., bounds=[-100., -50.]),
# Parameter('e_c_L', value=0., bounds=[-1./15, 1./15])]
elif c_name == 'L':
params = [Parameter('d_L', value=1./1000., bounds=[0.0, 0.05]),
Parameter('g0_L', value=0.40*1e2, bounds=[0., 300.]),
Parameter('e_0_L', value=-90., bounds=[-100., -50.])]
elif c_name == 'L_c':
params = [Parameter('d_L_c', value=1./1000., bounds=[0.0, 0.05]),
Parameter('g0_L_c', value=0.40*1e2, bounds=[0., 300.]),
Parameter('e_0_L_c', value=-90., bounds=[-100., -50.]),
Parameter('e_c_L_c', value=0., bounds=[-1./15, 1./15])]
# # original params
# self.params = [Parameter('d_c_m', value=0.001, bounds=[0., 0.01]),
# Parameter('c_m_0', value=1., bounds=[0.95, 1.05])]
# for c_name in self.channel_names:
# if c_name == 'K_m' or c_name == 'K_m35' or c_name == 'K_ir':
# params = [Parameter('d_'+c_name, value=-1./100., bounds=[-0.1, 0.]),
# Parameter('g0_'+c_name, value=40.*1e2, bounds=[0., 10000.]),
# Parameter('e_r_'+c_name, value=-85., bounds=[-95.,-80.])]
# elif c_name == 'h_HAY' or c_name == 'h_u':
# params = [Parameter('d_'+c_name, value=1./100., bounds=[0.0, 0.1]),
# Parameter('g0_'+c_name, value=0.0099*1e2, bounds=[0., 10000.]),
# Parameter('e_r_'+c_name, value=-40., bounds=[-50.,-30.])]
# elif c_name == 'Na_p' or c_name == 'NaP':
# params = [Parameter('d_'+c_name, value=0., bounds=[-0.0001, 0.0001]),
# Parameter('g0_'+c_name, value=0.01*1e2, bounds=[0.,10000.]),
# Parameter('e_r_'+c_name, value=50., bounds=[40.,60.])]
# # elif c_name == 'L':
# # params = [Parameter('d_L', value=1./200., bounds=[0.0, 0.05]),
# # Parameter('g0_L', value=0.40*1e2, bounds=[0., 300.]),
# # Parameter('e_0_L', value=-90., bounds=[-100., -50.]),
# # Parameter('e_c_L', value=0., bounds=[-1./15, 1./15])]
# elif c_name == 'L':
# params = [Parameter('d_L', value=1./200., bounds=[0.0, 0.05]),
# Parameter('g0_L', value=0.40*1e2, bounds=[0., 300.]),
# Parameter('e_0_L', value=-90., bounds=[-100., -50.])]
# elif c_name == 'L_c':
# params = [Parameter('d_L_c', value=1./200., bounds=[0.0, 0.05]),
# Parameter('g0_L_c', value=0.40*1e2, bounds=[0., 300.]),
# Parameter('e_0_L_c', value=-90., bounds=[-100., -50.]),
# Parameter('e_c_L_c', value=0., bounds=[-1./15, 1./15])]
else:
warnings.warn('unrecognized ion channel \'' + c_name +'\'( choose from ' + \
' '.join(['L', 'K_m', 'K_m35', 'K_ir', 'h_HAY', 'h_u']), + \
'), ignoring current channel.')
self.params.extend(params)
def evalFitness(self, responses):
return [obj.calculate_score(responses) for obj in self.objectives]
def getParameterValues(self):
return [p.value for p in self.params]
def setParameterValues(self, values):
if values is None:
values = self.getParameterValues()
if isinstance(values, list):
for p, v in zip(self.params, values): p.value = v
elif isinstance(values, dict):
for p in self.params: p.value = values[p.name]
else:
raise TypeError('``values`` must be `list` or `dict`')
def getParameterValuesAsDict(self):
return {p.name: p.value for p in self.params}
def toStrParameterValues(self, values=None):
rstr = 'Parametervalues =\n'
if values is None:
values = [p.value for p in self.params]
if isinstance(values, list):
for p, v in zip(self.params, values):
rstr += ' > ' + p.name + ' = %.5f\n'%v
elif isinstance(values, dict):
for p in self.params:
rstr += ' > ' + p.name + ' = %.5f\n'%values[p.name]
return rstr
def toStrFitness(self, responses):
fitness = self.evalFitness(responses)
rstr = 'Fitness =\n'
for ii, ff in enumerate(fitness):
rstr += ' > f_%d = %.5f\n'%(ii,ff)
return rstr
def getTreeWithParams(self, new_tree=None):
ps = self.getParameterValuesAsDict()
# set the physiology parameters of this tree
sim_tree = self.sim_tree.__copy__(new_tree=new_tree)
sim_tree.treetype = 'original'
# capacitance
c_m_distr = utils.linDistr(ps['c_m_0'], ps['d_c_m'])
sim_tree.setPhysiology(c_m_distr, self.r_a)
# membrane current parameters
for ii, c_name in enumerate(self.channel_names):
g_func = utils.expDistr(ps['d_'+c_name], ps['g0_'+c_name], g_max=self.g_max[c_name])
if c_name != 'L' and c_name != 'L_c':
e_r = ps['e_r_'+c_name]
# add the current
chan = eval('channelcollection.' + c_name + '()')
sim_tree.addCurrent(chan, g_func, e_r)
else:
if c_name == 'L_c':
e_func = utils.linDistr(ps['e_0_L_c'], ps['e_c_L_c'])
else:
e_func = lambda x: ps['e_0_L']
# add the current
for node in sim_tree:
d2s = sim_tree.pathLength({'node': node.index, 'x': .5}, (1., 0.5))
g_l = g_func(d2s)
e_l = e_func(d2s)
node._addCurrent('L', g_l, e_l)
return sim_tree
def runSim(self):
'''
Format for args:
[c_m, r_a] +
[d_scale, g_0, g_max] for expDistr for each conductance channel in self.channel_names +
[e_0, e_1] for the leak potential
'''
sim_tree = self.getTreeWithParams()
# initialize the simulation
sim_tree.setCompTree()
sim_tree.treetype = 'computational'
sim_tree.initModel(dt=self.dt, t_calibrate=200.)
# add Iclamps
sim_tree.addIClamp(self.loc_dend, self.a0, self.t0, self.dur)
sim_tree.addIClamp(self.loc_dend, self.a1, self.t1, self.dur)
sim_tree.addIClamp(self.loc_soma, self.a2, self.t2, self.dur)
sim_tree.addIClamp(self.loc_soma, self.a3, self.t3, self.dur)
# set recorders
sim_tree.storeLocs([self.loc_soma, self.loc_dend], name='rec locs')
# run simulation
res = sim_tree.run(self.t_max, pprint=False)
sim_tree.deleteModel()
return res
def evaluate_with_lists(self, param_values=None):
return self.evaluate(param_values)
def evaluate(self, param_values, pprint=True):
self.setParameterValues(values=param_values)
res = self.runSim()
fitness = self.evalFitness(res['v_m'][:,:-1])
if pprint:
print('>>> fitness =', fitness)
# print '>>> ' + self.toStrParameterValues()
return fitness
class AttenuationEvaluator(ModelEvaluator):
def __init__(self, sim_tree, f_d2s, f_s2d,
loc_soma, loc_dend, mode='fit'):
"""
Only optimizes h-current
if mode is fit, check bounds, if mode is evaluate, dont check bounds
"""
self.sim_tree = sim_tree
# injection sites
self.loc_soma, self.loc_dend = loc_soma, loc_dend
# params
self.dt, self.dur, self.t_max = data.DT, data.DUR, data.T_MAX
self.t0, self.t1, self.t2, self.t3 = data.T0, data.T1, data.T2, data.T3
self.a0, self.a1, self.a2, self.a3 = data.A0, data.A1, data.A2, data.A3
self.mode = mode
# define fit parameters
self._defineFitObjects(f_d2s, f_s2d)
# don't check bounds
if mode == 'evaluate':
for p in self.params:
p.bounds = None
def _defineFitObjects(self, f_d2s, f_s2d):
# reference attenuation
v_dat = data.DataContainer(with_zd=True)
att_f = utils.AttFeature(v_dat)
att_ref_d2s = att_f.att_d2s * f_d2s
att_ref_s2d = att_f.att_s2d * f_d2s
# fitness evaluator
features = [utils.AttFeature_(att_ref_d2s, att_ref_s2d, v_dat)]
self.objectives = [SingletonObjective('Att', features[0])]
# parameters
self.params = [
Parameter('g_h_0', value=0., bounds=[0.,50000.]),
Parameter('g_h_1', value=200., bounds=[0.,50000.]),
Parameter('g_h_2', value=2000., bounds=[0.,50000.]),
Parameter('g_h_3', value=5000., bounds=[0.,50000.]),
Parameter('e_r_h', value=-40., bounds=[-50.,-30.]),
Parameter('g_h_b', value=0., bounds=[0.,50000.]),
]
def _h_distr_func(self, x, ds=[0., 250., 500., 750.]):
ps = self.getParameterValues()
if x <= ds[1]:
d0 = ds[0]; d1 = ds[1]
p0 = ps[0]; p1 = ps[1]
elif x > ds[1] and x < ds[2]:
d0 = ds[1]; d1 = ds[2]
p0 = ps[1]; p1 = ps[2]
else:
d0 = ds[2]; d1 = ds[3]
p0 = ps[2]; p1 = ps[3]
return p0 + (p1 - p0) / (d1 - d0) * (x - d0)
def getTreeWithParams(self, new_tree=None):
ps = self.getParameterValuesAsDict()
# set the physiology parameters of this tree
sim_tree = self.sim_tree.__copy__(new_tree=new_tree)
sim_tree.treetype = 'original'
# h-current distribution
h_u = channelcollection.h_u()
sim_tree.addCurrent(h_u, self._h_distr_func, ps['e_r_h'], node_arg='apical')
sim_tree.addCurrent(h_u, ps['g_h_b'], ps['e_r_h'], node_arg='basal')
sim_tree.addCurrent(h_u, ps['g_h_b'], ps['e_r_h'], node_arg=[sim_tree[1]])
return sim_tree
################################################################################
def optimize(evaluator):
global MAX_ITER, N_OFFSPRING
optimisation = IBEADEAPOptimisation(evaluator=evaluator,
offspring_size=N_OFFSPRING, map_function=map)
final_pop, hall_of_fame, logs, hist = optimisation.run(max_ngen=MAX_ITER)
return final_pop, hall_of_fame, logs, hist
def optimizeModel(channel_names=None, zd=False, suffix=''):
"""
Optimizes the morphology equipped with channels in `channel_names` to
recordings with or without ZD
Parameters
----------
channel_names: list of str
Choose channel names from from {'L', 'K_m', 'K_m35', 'K_ir', 'h_HAY', 'h_u'}
zd: bool
True for data with ZD, false for data without ZD
"""
global MAX_ITER, N_OFFSPRING
if channel_names is None:
channel_names = ['L', 'K_ir', 'K_m35', 'h_u']
file_name = utils.getFileName(channel_names, zd, suffix=suffix)
full_tree, red_tree, full_locs, red_locs = data.reduceMorphology()
sim_tree = red_tree.__copy__(new_tree=NeuronSimTree())
# measured data
v_dat = data.DataContainer(with_zd=zd)
model_evaluator = ModelEvaluator(sim_tree, v_dat, red_locs[0], red_locs[1],
channel_names=channel_names)
final_pop, hall_of_fame, logs, hist = optimize(model_evaluator)
# save hall of fame
file = open(file_name, 'wb')
pickle.dump(hall_of_fame, file)
file.close()
if __name__ == "__main__":
optimizeModel(channel_names=['L'], zd=False, suffix='_test')