def simulate_diff_3x_freq(hzs=[10,20, 30],n_msns=[40/2, 20/2, 13/2], res=20, load=True):
save_result_at=OUTPUT_PATH+'/simulate_diff_3x_freq.pkl'
n_exp=100
if not load:
spk_mean=[]
for hz, n_msn in zip(hzs,n_msns):
sim_time= SEL_ONSET+1000
params_msn={'base_rates':[MSN_BASE_RATE], 'base_times':[1], 'mod_rates': [MSN_BASE_RATE, hz, MSN_BASE_RATE],
'mod_times':[1,SEL_ONSET, SEL_ONSET+500], 'n_tot':N_MSN, 'n_mod':n_msn}
params_gpe={'base_rates':[GPE_BASE_RATE], 'base_times':[1], 'n_tot':N_GPE, 'n_mod':0}
params_stn={'base_rates':[STN_BASE_RATE], 'base_times':[1], 'n_tot':N_STN, 'n_mod':0}
synapse_models={'MSN':'MSN_SNR_gaba_p1', 'GPE':'GPE_SNR_gaba_p',
'STN':'STN_SNR_ampa_s'}
#times, spk_binned =_simulate_model([params_msn, params_gpe,params_gpe, 'SNR_izh',
# synapse_models, sim_time, 0])
t=ttime.time()
times, spk_binned =simulate_model(params_msn, params_gpe, params_stn, 'SNR_izh',
synapse_models, sim_time, res, n_exp=n_exp,threads=4)
print 'Time:',ttime.time()-t
spk_mean.append(numpy.mean(spk_binned,axis=0))
spk_mean=numpy.array(spk_mean)*1000/res
misc.pickle_save([times, spk_mean], save_result_at)
else:
times, spk_mean = misc.pickle_load(save_result_at)
return times, spk_mean
def fmin(load, save_at):
x0=[188, 0.08] #[current, w_GPE_STN] 20 Hz
#29.720625
#0.011041875
# x0=[290, 0.119] #25
# x0=[430,0.18] #30 Hz
# x=[540, 0.215] #35 Hz
# x0=[702, 0.28] #40 Hz
# x0=[830., 0.336] #45 Hz
# x0=[876.7, 0.349] #[current, w_GPE_STN] 46 Hz
# x0=[1000.8, 0.3957] # 50 Hz]
# x0=[1159., 0.458] # 55 Hz]
# x0=[1159.+2.5*5*29.7, 0.458+2.5*5*0.01104] # 80 Hz]
x0=[2102, 0.794] # 80 Hz]
# z=[1161, 454] #
if not load:
[xopt,fopt, iter, funcalls , warnflag, allvecs] = opt.fmin(error_fun,
x0,
args=([sim_time]),
maxiter=20,
maxfun=20,
full_output=1, retall=1)
misc.pickle_save([xopt,fopt, iter, funcalls , warnflag, allvecs], save_at)
else:
[xopt,fopt, iter, funcalls , warnflag, allvecs]=misc.pickle_load(save_at)
return xopt
def simulate_filterting_burst(load, save_at, interval, N_MSN, params_msn_d1,
params_msn_d2, models_msn, sim_time, start_re):
N_MSN_syn_on_SNR = 500.
max_base_rate = 0.9
max_syn_events_burst = 1000 # Max for burst, then need to add on the contribution of
# background, See at return statement.
msn_burst_rate = 20.
msn_base_rate = 0.1
n_max_bursting = (max_syn_events_burst) / msn_burst_rate - 1
n_burst_per_SNR = numpy.arange(1, n_max_bursting, 2)
prop_mod = n_burst_per_SNR / N_MSN_syn_on_SNR
mod_const_syn = prop_mod * N_MSN
mod = prop_mod * float(N_MSN)
mod = numpy.array([int(m) for m in mod])
freq = numpy.ones(len(mod)) * msn_burst_rate
syn_events_burst = numpy.array(mod * freq +
(N_MSN - mod) * 0.1) * 500. / N_MSN
model_params = {
'misc': {
'N_MSN': N_MSN
},
'conns': {
'MSN_D2_GPE': {
'lines': False
}
},
'neurons': {
'MSN_D1': {
'n': N_MSN
},
'MSN_D2': {
'n': N_MSN
}
}
}
seed = range(len(mod))
base_rates = numpy.ones(len(mod)) * 0.1
if not load:
mr = simulate_filterting_burst_fun(mod, freq, base_rates,
params_msn_d1, params_msn_d2,
params_stn, synapse_models,
sim_time, seed, {}, threads,
start_rec, model_params)
misc.pickle_save(mr, save_at)
else:
mr = misc.pickle_load(save_at)
mrb = numpy.mean(mr[:, interval[0]:interval[1]], axis=1)
syn_ev_b = numpy.array(mod * freq + (N_MSN - mod) * 0.1) * 500. / N_MSN
data = numpy.array([syn_ev_b, mrb])
return data
def simulate(load,
save_at,
n_exp,
res,
params_msn_d1,
params_msn_d2,
params_stn,
synapse_models,
sim_time,
seed,
start_rec,
I_e_add,
threads,
model_params,
dis_conn_GPE_STN=False):
if not load:
r = []
for i in range(n_exp):
seed = i
p = numpy.random.randint(1)
start_rec += p
sim_time += p
layer_dic = simulate_network(params_msn_d1,
params_msn_d2,
params_stn,
synapse_models,
sim_time,
seed,
I_e_add,
threads,
start_rec,
model_params,
dis_conn_GPE_STN=dis_conn_GPE_STN)
layer_dic['SNR'].get_signal('s', start=start_rec, stop=sim_time)
signal = layer_dic['SNR'].signals['spikes']
r.append(
numpy.mean(signal.spike_histogram(time_bin=1, normalized=True),
axis=0))
numpy.array(r)
misc.pickle_save(r, save_at)
else:
r = misc.pickle_load(save_at)
r = numpy.array(r)
r = misc.convolve(r, res, 'triangle', single=False)
mr = numpy.mean(r, axis=0)
mstd = numpy.std(r, axis=0)
d = [mr, mstd]
return d
def my_fmin(load, save_at, x0, interval, syn, N_MSN, burst_rate):
x = x0
if not load:
x, e, i, allvecs = my_opt(x,
sim_time,
interval,
syn,
burst_rate,
N_MSN,
maxiter=10)
misc.pickle_save([x, e, i, allvecs], save_at)
else:
[x, e, i, allvecs] = misc.pickle_load(save_at)
return x, e
def fmin(load, save_at, x0, n_exp, n_gpe, res1_stn_rate_ch):
sim_time = 10000
if not load:
[xopt, fopt, iter, funcalls, warnflag,
allvecs] = opt.fmin(error_fun,
x0,
args=([sim_time, n_exp, n_gpe, res1_stn_rate_ch]),
maxiter=10,
maxfun=10,
full_output=1,
retall=1)
misc.pickle_save([xopt, fopt, iter, funcalls, warnflag, allvecs],
save_at)
else:
[xopt, fopt, iter, funcalls, warnflag,
allvecs] = misc.pickle_load(save_at)
return xopt, fopt
def simulate_filtering(load, save_at, N_MSN, params_msn_d1, params_msn_d2,
models_msn, sim_time, start_rec):
freq_filter = numpy.linspace(0.1, 2.6, 10)
seed = range(len(freq_filter))
model_params = {
'misc': {
'N_MSN': N_MSN
},
'neurons': {
'MSN_D1': {
'n': N_MSN
},
'MSN_D2': {
'n': N_MSN
}
}
}
if not load:
mr = []
for syn in models_msn:
synapse_models = [syn, 'GPE_SNR_gaba_p']
mr.append(
simulate_filtering_fun(freq_filter, params_msn_d1,
params_msn_d2, params_stn,
synapse_models, sim_time, seed, {},
threads, start_rec, model_params))
mr = numpy.array(mr)
misc.pickle_save(mr, save_at)
else:
mr = misc.pickle_load(save_at)
syn_ev = freq_filter * N_MSN * 500. / N_MSN
# Row one in mr is s min, then s max and finally the plastic synapse
data = numpy.array([syn_ev, freq_filter, mr[0, :], mr[1, :], mr[2, :]])
return data
def fmin(load, save_at, x0, interval, syn, N_MSN, burst_rate):
#[current, w_GPE_STN]
args = (sim_time, interval, syn, burst_rate, N_MSN)
if not load:
[xopt, fopt, iter, funcalls, warnflag,
allvecs] = opt.fmin(error_fun,
x0,
args=args,
maxiter=20,
maxfun=10,
full_output=1,
retall=1)
misc.pickle_save([xopt, fopt, iter, funcalls, warnflag, allvecs],
save_at)
else:
[xopt, fopt, iter, funcalls, warnflag,
allvecs] = misc.pickle_load(save_at)
return xopt, fopt
def fmin(load, save_at):
sim_time = 10000
x0 = [42, 1.3, 0.35] #[50, 0.9, 0.2] #[current, w_GPE_GPE, w_STN_GPE]
if not load:
[xopt, fopt, iter, funcalls, warnflag,
allvecs] = opt.fmin(error_fun,
x0,
args=([sim_time]),
maxiter=50,
maxfun=50,
full_output=1,
retall=1)
misc.pickle_save([xopt, fopt, iter, funcalls, warnflag, allvecs],
save_at)
else:
[xopt, fopt, iter, funcalls, warnflag,
allvecs] = misc.pickle_load(save_at)
return xopt
def simulate_signal_rates(load=True, hzs=[1, 2]):
# Path were raw data is saved. For example the spike trains.
save_result_at = DP['OUTPUT_PATH'] + '/simulate_signal_rates.pkl'
save_header_at = DP['OUTPUT_PATH'] + '/simulate_signal_rates_header'
rates = []
if not load:
for hz in hzs:
rates.append(simulate_get_rates(msn_burst_rate=hz, load=load))
rates = numpy.array(rates)
header = HEADER_SIMULATION_SETUP
misc.text_save(header, save_header_at)
misc.pickle_save(rates, save_result_at)
else:
rates = misc.pickle_load(save_result_at)
return rates
def simulate_1500_eval(load, save_at, n_exp, params_msn_d1, params_msn_d2,
params_stn, synapse_models, sim_time, seed, I_e_add,
threads, start_rec):
rates = []
if not load:
for i in range(n_exp):
seed = i
layer_dic, r = simulate_1500(params_msn_d1, params_msn_d2,
params_stn, synapse_models, sim_time,
seed, I_e_add, threads, start_rec)
rates.append(r)
rates = numpy.array(rates)
misc.pickle_save(rates, save_at)
else:
rates = misc.pickle_load(save_at)
mr = numpy.mean(rates, axis=0)
stdr = numpy.std(rates, axis=0)
return mr, stdr
def fmin(load, save_at, x0, n_exp, r_target, params_msn_d1, params_msn_d2,
params_stn, sim_time, I_e_add, threads, start_rec, model_params,
p_weights):
#[current, w_GPE_STN]
args = (n_exp, r_target, params_msn_d1, params_msn_d2, params_stn,
sim_time, I_e_add, threads, start_rec, model_params, p_weights)
if not load:
[xopt, fopt, iter, funcalls, warnflag,
allvecs] = opt.fmin(error_fun,
x0,
args=args,
maxiter=20,
maxfun=20,
full_output=1,
retall=1)
misc.pickle_save([xopt, fopt, iter, funcalls, warnflag, allvecs],
save_at)
else:
[xopt, fopt, iter, funcalls, warnflag,
allvecs] = misc.pickle_load(save_at)
return xopt, fopt
def simulate_rate_first_and_second_bursts_full(load=True):
global OUTPUT_PATH
save_result_at = OUTPUT_PATH + '/simulate_rate_first_and_second_bursts_full.pkl'
save_header_at = OUTPUT_PATH + '/simulate_rate_first_and_second_bursts_full_header'
# Range
transient_stops = numpy.arange(100, 3200, 500)
#hzs=[8,20]
if not load:
data = {}
data['rates'] = []
for stop in transient_stops:
mean_rates, mean_rates_std, info_string = simulate_rate_first_and_second_bursts(
selection_intervals=[0.0, 500.0, 500. + stop, 1000. + stop],
load=False)
data['rates'].append(mean_rates[0])
s = '\n'
s = s + 'simulate_rate_first_and_second_bursts_full\n'
s = s + ' %s %5s %s \n' % ('Transient stops', str(transient_stops[0]) +
'-' + str(transient_stops[-1]), 'ms')
header = s
misc.text_save(header, save_header_at)
misc.pickle_save([data, s], save_result_at)
info_string = s
elif load:
data, info_string = misc.pickle_load(save_result_at)
data['rates'] = numpy.array(data['rates'])
return transient_stops, data, info_string
synapse_models = [model, 'GPE_SNR_gaba_p']
layer_dics.append(
simulate_network(params_msn_d1,
params_msn_d2,
params_stn,
synapse_models,
sim_time=sim_time,
seed=seed,
I_e_add={
'SNR': 300,
'STN': 0,
'GPE': 30
},
threads=4,
start_rec=500.))
misc.pickle_save(layer_dics, save_result_at)
else:
layer_dics = misc.pickle_load(save_result_at)
params_msn_d1 = {
'base_rates': [0.1],
'base_times': [1],
'mod_rates': [0.1, 20, 0.1, 20, 0.1],
'mod_times': [1, 1000, 1000 + 500],
'n_mod': 60
}
params_msn_d2 = {
'base_rates': [0.1],
'base_times': [1],
'mod_rates': [0.1, 20, 0.1],
'mod_times': [1, 1000, 1000 + 500],
def simulate_sensitivity(load, save_result_at, n_exp, params_msn_d1, params_msn_d2,
params_stn, synapse_models,model_params, sim_time,
start_rec):
p_weights=numpy.ones(17)
p_weights_ch_names=[r'$g^{CTX-STN}_{0}$','$g^{GPE-STN}_{0}$','$g^{MSN_{D2}-GPe}_0$',
'$g^{STN-GPe}_{0}$','$g^{GPe-GPe}_{0}$','$g^{MSN_{D1}-SNr}_0$',
'$g^{STN-SNr}_{0}$','$g^{GPe-SNr}_{0}$']
#p_weights_ch_names=['$g^{STN-SNr}_{0}$','$g^{GPe-SNr}_{0}$']
p_weights_ch=[6, 8, 9, 10, 12, 13, 14, 16]
#p_weights_ch=[14,16]
p_conn=numpy.ones(7)
p_conn_names=[r'$N_{MSN_{D1}-SNr}$', r'$N_{MSN_{D2}-GPe}$', r'$N_{STN-GPe}$',
r'$N_{GPe-GPe}$', r'$N_{GPe-STN}$', r'$N_{STN-SNr}$',
r'$N_{GPe-SNr}$']
#p_conn_names=[r'$N_{STN-SNr}$', r'$N_{GPe-SNr}$']
p_conn_ch=[0, 1, 2, 3, 4 ,5, 6]
p_I_e=numpy.ones(3)
#p_I_e_names=['$I_{In vivo}^{SNr}$', '$I_{In vivo}^{GPe}$', '$I_{In vivo}^{STN}$']
#p_I_e_ch=[0,1,2]
d=[]
prop_ch=0.2 # Change each parameter up and down 0.2
seed=0
save_result_at_tmp=save_result_at+'weights'
if not load[0]:
dd=[]
seed=2
mr, mstd=simulate_sensitivity_fun(n_exp, params_msn_d1, params_msn_d2, params_stn,
synapse_models, sim_time, seed,
{}, threads,
start_rec, model_params, p_weights, p_conn, p_I_e)
dd=[[-1,0,mr[0],mr[1],mr[2], mstd[0],mstd[1],mstd[2], 0]]
for p in p_weights_ch:
for prop in [1+prop_ch,1-prop_ch]:
seed+=1
p_weights_tmp=copy.deepcopy(p_weights)
p_weights_tmp[p]*=prop
mr, mstd=simulate_sensitivity_fun(n_exp,params_msn_d1, params_msn_d2, params_stn,
synapse_models, sim_time, seed,
{}, threads,
start_rec,model_params, p_weights_tmp, p_conn, p_I_e)
dd.append([ 0,0, mr[0], mr[1], mr[2], mstd[0], mstd[1], mstd[2], 0])
misc.pickle_save(dd,save_result_at_tmp)
else:
dd=misc.pickle_load(save_result_at_tmp)
d.extend(dd)
save_result_at_tmp=save_result_at+'conn'
if not load[1]:
dd=[]
for p in p_conn_ch:
for prop in [1+prop_ch,1-prop_ch]:
seed+=1
p_conn_tmp=copy.deepcopy(p_conn)
p_conn_tmp[p]*=prop
mr, mstd=simulate_sensitivity_fun(n_exp,params_msn_d1, params_msn_d2, params_stn,
synapse_models, sim_time, seed,
{}, threads,
start_rec, model_params, p_weights, p_conn_tmp, p_I_e)
dd.append([1,0,mr[0],mr[1],mr[2], mstd[0],mstd[1],mstd[2], 0])
misc.pickle_save(dd, save_result_at_tmp)
else:
dd=misc.pickle_load(save_result_at_tmp)
d.extend(dd)
# save_result_at_tmp=save_result_at+'I_e'
# if not load[2]:
# d=[]
# for p in p_I_e_ch:
# for prop in [1+prop_ch,1-prop_ch]:
# seed+=1
# p_I_e_tmp=copy.deepcopy(p_I_e)
# p_I_e_tmp[p]*=prop
#
# mr, mstd=simulate_sensitivity_fun(n_exp,params_msn_d1, params_msn_d2, params_stn,
# synapse_models, sim_time, seed,
# {}, threads, start_rec, model_params,
# p_weights, p_conn, p_I_e_tmp)
#
# d.append([2,0,mr[0],mr[1],mr[2], mstd[0],mstd[1],mstd[2], 0])
#
# misc.pickle_save(d,save_result_at_tmp)
# else:
# d.extend(misc.pickle_load(save_result_at_tmp) )
d=numpy.array(d)
br=d[0,2:5]
bstd=d[0,5:8]
up=d[1::2,2:5]
down=d[2::2,2:5]
upstd=d[1::2,5:8]
downstd=d[2::2,5:8]
dp=numpy.abs(up-down)/2.
# get as precent change
for i in range(up.shape[0]):
up[i,:]-=br
up[i,:]/=br
up[i,:]*=100.
down[i,:]-=br
down[i,:]/=br
down[i,:]*=100.
upstd[i,:]-=bstd
upstd[i,:]/=bstd
upstd[i,:]*=100.
downstd[i,:]-=bstd
downstd[i,:]/=bstd
downstd[i,:]*=100.
data_rate_change=[[up, upstd],[down, downstd]]
return numpy.array(d), data_rate_change, dp, p_weights_ch_names + p_conn_names #+ p_I_e_names
def simulate_recovery(revoceryTimes, load=True):
# Path were raw data is saved. For example the spike trains.
save_result_at=OUTPUT_PATH+'/simulate_recovery.pkl'
save_header_at=OUTPUT_PATH+'/simulate_recovery_header'
relativeRecovery=[]
n=len(revoceryTimes)
if not load:
for syn in SYNAPSE_MODELS:
my_nest.ResetKernel()
model_list, model_dict=models()
my_nest.MyLoadModels( model_list, NEURON_MODELS )
my_nest.MyLoadModels( model_list, [syn])
ss=my_nest.GetDefaults(syn)
synapticEficacy = ss['weight']*ss['U']
SNR = MyGroup( NEURON_MODELS[0], n, mm=True, mm_dt = .1,
params={'I_e':-150.}, record_from=['g_AMPA'])
tSim=10000
spikeTimes=[]
for rt in revoceryTimes:
#spikeTimes.append(numpy.array([1.,11.,21.,31.,41.,41+rt]))
# Choosen so that it starts at a pairpulse ration of 0.2
spikeTimes.append(numpy.array([1.,11.,21.,31.,41.,
51.,61.,71.,81.,91.,
101.,111.,121.,131.,141.,
151.,161.,171.,181.,191.,
191+rt]))
for target, st in zip(SNR, spikeTimes ) :
source = my_nest.Create('spike_generator',
params={'spike_times':st} )
my_nest.SetDefaults(syn, params={'delay':1.})
my_nest.Connect(source, [target], model=syn)
my_nest.MySimulate(tSim)
SNR.get_signal( 'g','g_AMPA', stop=tSim ) # retrieve signal
signal=SNR.signals['g_AMPA']
tmpSteadyState=[]
for i, st in enumerate(spikeTimes, start=1):
if SNR.mm_dt==0.1: indecies=numpy.int64(numpy.ceil(st*10))+9
elif SNR.mm_dt==1.: indecies=numpy.int64(numpy.ceil(st))
values=signal[i].signal[indecies]-signal[i].signal[indecies-1]
tmpSteadyState.append(values[-1]/synapticEficacy)
#tmpSteadyState.append(max(values)/synapticEficacy)
relativeRecovery.append(tmpSteadyState)
relativeRecovery=numpy.array(relativeRecovery)
header=HEADER_SIMULATION_SETUP
misc.text_save(header, save_header_at)
misc.pickle_save([revoceryTimes, relativeRecovery], save_result_at)
elif load:
revoceryTimes, relativeRecovery=misc.pickle_load(save_result_at)
return revoceryTimes, relativeRecovery
def simulate_steady_state_freq(frequencies, flag='ss', load=True):
# Path were raw data is saved. For example the spike trains.
save_result_at=OUTPUT_PATH+'/simulate_steady_state_freq.pkl'
save_header_at=OUTPUT_PATH+'/simulate_steady_state_freq_header'
relativeFacilitation=[]
n=len(frequencies)
if not load:
for syn in SYNAPSE_MODELS:
my_nest.ResetKernel()
model_list, model_dict=models()
my_nest.MyLoadModels( model_list, NEURON_MODELS )
my_nest.MyLoadModels( model_list, [syn])
SNR = MyGroup( NEURON_MODELS[0], n, mm=True, mm_dt = .1,
params={'I_e':-150.},
record_from=['g_AMPA'] )
tSim=5*1000/frequencies[0]
spikeTimes=[]
tmpSteadyState=[]
for f in frequencies :
isi = 1000./f
spikeTimes.append(numpy.arange(1,tSim,isi))
for target, st in zip(SNR, spikeTimes ) :
source = my_nest.Create('spike_generator',
params={'spike_times':st} )
my_nest.SetDefaults(syn, params={'delay':1.})
my_nest.Connect(source, [target], model=syn)
my_nest.MySimulate(tSim)
SNR.get_signal( 'g','g_AMPA', stop=tSim ) # retrieve signal
signal=SNR.signals['g_AMPA']
for i, st in enumerate(spikeTimes, start=1):
if SNR.mm_dt==0.1: indecies=numpy.int64(numpy.ceil(st*10))+9
elif SNR.mm_dt==1.: indecies=numpy.int64(numpy.ceil(st))
values=signal[i].signal[indecies]-signal[i].signal[indecies-1]
ss=my_nest.GetDefaults(syn)
synapticEficacy = ss['weight']*ss['U']
if flag=='ss': tmpSteadyState.append(values[-1]/synapticEficacy)
if flag=='max': tmpSteadyState.append(max(values)/synapticEficacy)
relativeFacilitation.append(tmpSteadyState)
relativeFacilitation=numpy.array(relativeFacilitation)
header=HEADER_SIMULATION_SETUP
misc.text_save(header, save_header_at)
misc.pickle_save([frequencies, relativeFacilitation], save_result_at)
elif load:
frequencies, relativeFacilitation=misc.pickle_load(save_result_at)
return frequencies, relativeFacilitation
def simulate_example(hz=0, load=True):
global SNR_INJECTED_CURRENT
global NEURON_MODELS
global N_GPE
global N_SEL
global N_MSN
global N_STN
global MSN_RATE_BASE
global STN_BASE_RATE
global SNAME
global SPATH
global SYNAPSE_MODELS
global SEL_ONSET
global GPE_BASE_RATE
#n_exp = 20
n_exp = 200
RATE_SELE = hz # Selection rate
save_at = SPATH + '/' + NEURON_MODELS[0] + '-example.pkl'
sim_time = SEL_TIME + SEL_ONSET + 500.
SNAME_NB = hz + 1000
experiments = range(n_exp)
MODEL_LIST = models()
my_nest.ResetKernel()
my_nest.MyLoadModels(MODEL_LIST, NEURON_MODELS)
my_nest.MyLoadModels(MODEL_LIST, SYNAPSE_MODELS_TESTED)
my_nest.MyLoadModels(MODEL_LIST, SYNAPSE_MODELS_BACKGROUND)
GPE_list = [] # GPE input for each experiment
for i_exp in experiments:
GPE = MyPoissonInput(n=N_GPE,
sd=True,
spath=SPATH,
sname_nb=SNAME_NB + i_exp)
GPE_list.append(GPE)
MSN_list = [] # MSN input for each experiment
for i_exp in experiments:
MSN = MyPoissonInput(n=N_MSN, sd=False)
MSN_list.append(MSN)
STN_list = [] # MSN input for each experiment
for i_exp in experiments:
STN = MyPoissonInput(n=N_STN, sd=False)
STN_list.append(STN)
SNR_list = [] # SNR groups for each synapse
I_e = my_nest.GetDefaults(NEURON_MODELS[0])['I_e'] + SNR_INJECTED_CURRENT
for i_syn in range(len(SYNAPSE_MODELS_TESTED)):
SNR = MyGroup(NEURON_MODELS[0], n=n_exp, params={'I_e': I_e}, sd=True)
SNR_list.append(SNR)
if not load:
for i_exp in experiments:
GPE = GPE_list[i_exp]
MSN = MSN_list[i_exp]
STN = STN_list[i_exp]
# Set spike times
# Base rate MSN
for id in MSN[:]:
MSN.set_spike_times(id=id,
rates=[MSN_RATE_BASE],
times=[1],
t_stop=sim_time,
seed=int(numpy.random.random() * 10000.0))
# Base rate STN
for id in STN[:]:
STN.set_spike_times(id=id,
rates=[STN_BASE_RATE],
times=[1],
t_stop=sim_time,
seed=int(numpy.random.random() * 10000.0))
# Set spike times
# Base rate
for id in GPE[0:N_GPE - N_SEL]:
GPE.set_spike_times(id=id,
rates=[GPE_BASE_RATE],
times=[1],
t_stop=sim_time,
seed=int(numpy.random.random() * 10000.0))
# Selection
for id in GPE[N_GPE - N_SEL:N_GPE]:
rates = [GPE_BASE_RATE, RATE_SELE, GPE_BASE_RATE]
times = [1, SEL_ONSET, SEL_TIME + SEL_ONSET]
t_stop = sim_time
GPE.set_spike_times(id=id,
rates=rates,
times=times,
t_stop=t_stop,
seed=int(numpy.random.random() * 10000.0))
# Connect
for i_syn, syn in enumerate(SYNAPSE_MODELS_TESTED):
target = SNR_list[i_syn][i_exp]
my_nest.ConvergentConnect(GPE[:], [target], model=syn)
my_nest.ConvergentConnect(MSN[:], [target],
model=SYNAPSE_MODELS_BACKGROUND[0])
my_nest.ConvergentConnect(STN[:], [target],
model=SYNAPSE_MODELS_BACKGROUND[1])
my_nest.MySimulate(sim_time)
for GPE in GPE_list:
GPE.get_signal('s')
for SNR in SNR_list:
SNR.get_signal('s')
misc.pickle_save([GPE_list, SNR_list], save_at)
elif load:
GPE_list, SNR_list = misc.pickle_load(save_at)
pre_ref = str(SNR_list[0].signals['spikes'].mean_rate(
SEL_ONSET - 5000, SEL_ONSET))
pre_dyn = str(SNR_list[1].signals['spikes'].mean_rate(
SEL_ONSET - 500, SEL_ONSET))
statusSynapes = []
for syn in SYNAPSE_MODELS_TESTED:
statusSynapes.append(my_nest.GetDefaults(syn))
s = '\n'
s = s + 'Example:\n'
s = s + ' %s %5s %3s \n' % ('N experiments:', str(len(experiments)), '#')
s = s + ' %s %5s %3s \n' % ('N GPEs:', str(N_GPE), '#')
s = s + ' %s %5s %3s \n' % ('Base rate:', str(GPE_BASE_RATE), 'spikes/s')
s = s + ' %s %5s %3s \n' % ('Selection rate:', str(RATE_SELE), 'spikes/s')
s = s + ' %s %5s %3s \n' % ('Selection time:', str(SEL_TIME), 'ms')
s = s + ' %s %5s %3s \n' % ('Pre sel rate Ref:', pre_ref[0:4], 'spikes/s')
s = s + ' %s %5s %3s \n' % ('Pre sel rate Dyn:', pre_dyn[0:4], 'spikes/s')
for ss in statusSynapes:
s = s + '\n'
s = s + ' %s %10s\n' % ('Synapse', ss['synapsemodel'])
s = s + ' %s %5s %3s\n' % ('Weight', str(round(ss['weight'], 1)), 'nS')
return GPE_list, SNR_list, s
def simulate_selection_vs_neurons_full(selRateInterval,
load_pickle=True,
load_raw=True):
global OUTPUT_PATH
save_result_at = OUTPUT_PATH + '/simulate_selection_vs_neurons_full.pkl'
save_header_at = OUTPUT_PATH + '/simulate_selection_vs_neurons_full_header'
# Range
hzs = numpy.arange(7, 49, 1)
#hzs=[8,20]
if not load_pickle:
data = {}
for syn in range(3):
data[syn] = {}
data[syn]['rates_thr'] = [[] for k in range(len(SEL_INTERVALS))]
data[syn]['rates_std_thr'] = [[]
for k in range(len(SEL_INTERVALS))]
data[syn]['msn_at_thr'] = [[] for k in range(len(SEL_INTERVALS))]
data[syn]['n_max_sel'] = [[] for k in range(len(SEL_INTERVALS))]
n_max_sel = 218
progress = ''
i_hz = 0
for hz in hzs:
n, rate_data, r_std_data, n_max_sel, s = simulate_selection_vs_neurons(
SEL_INTERVALS, hz, load_raw, n_max_sel=n_max_sel)
n_sel_vec = numpy.arange(n_max_sel + 1)
# Clear n_max_cell
n_max_sel = 0
for i_interval in range(len(rate_data)):
for i_syn in range(len(rate_data[i_interval])):
r_syn = rate_data[i_interval][i_syn]
r_std_syn = r_std_data[i_interval][i_syn]
# Retrieve selection threshold passing
r_std_syn_tmp = r_std_syn[r_syn < SELECTION_THR]
n_sel_vec_tmp = n_sel_vec[r_syn < SELECTION_THR]
r_syn_tmp = r_syn[r_syn < SELECTION_THR]
data[i_syn]['rates_thr'][i_interval].append(r_syn_tmp[0])
data[i_syn]['rates_std_thr'][i_interval].append(
r_std_syn_tmp[0])
data[i_syn]['msn_at_thr'][i_interval].append(
n_sel_vec_tmp[0])
# Find new n_max_sel
msn_at_thr = data[i_syn]['msn_at_thr'][i_interval][i_hz]
n_max_sel = int(
numpy.ceil(max(msn_at_thr * 2.0, n_max_sel)))
data[i_syn]['n_max_sel'][i_interval].append(n_max_sel)
i_hz += 1
progress += str(hz) + ' hz finished, n_max_sel=' + str(
n_max_sel) + '\n'
print progress
s = '\n'
s = s + 'simulate_selection_vs_neurons_full\n'
s = s + ' %s %5s %s \n' % ('Range hz',
str(hzs[0]) + '-' + str(hzs[-1]), '#')
header = HEADER_SIMULATION_SETUP + s
misc.text_save(header, save_header_at)
misc.pickle_save([data, s], save_result_at)
info_string = s
elif load_pickle:
data, info_string = misc.pickle_load(save_result_at)
return hzs, data, info_string
def simulate_hyperdirect(load, N_MSN, save_at, threads, flag_bg=False):
ra = random.random() * 200.
start_rec = 900 + ra
delay = 10.0
params_msn_d1 = {
'base_rates': [0.1],
'base_times': [1],
'mod_rates': [0.1, 5000.0, 0.1],
'mod_times': [1, 1000. + ra + delay, 1000. + ra + 2. + delay],
'n_mod': 0.,
'bg_rate': 0
}
params_msn_d2 = {
'base_rates': [0.1],
'base_times': [1],
'mod_rates': [0.1, 0.1, 0.1],
'mod_times': [1, 1000, 1000 + 500],
'n_mod': 0,
'focus': False,
'skip': 1,
'bg_rate': 0
}
params_stn = {
'rate': 300.,
'mod': True,
'mod_rate': 0.,
'mod_times': [1000. + ra, 1000. + ra + 2.]
}
model_params = {
'conns': {
'MSN_D2_GPE': {
'lines': False
},
'STN_GPE': {
'lesion': True
},
'GPE_STN': {
'lesion': True
}
},
'neurons': {
'MSN_D1': {
'n': N_MSN
},
'MSN_D2': {
'n': N_MSN
}
}
}
synapse_models = ['MSN_SNR_gaba_p1', 'GPE_SNR_gaba_p']
if flag_bg: save_at = save_at + '_bg'
sim_time = 1300.
resolution = 5
n_exp = 5
proportions_d1 = numpy.linspace(0.01, 0.1, resolution) #arange(1,7,1)*150.
proportions_d2 = numpy.linspace(0.01, 0.1, resolution) #arange(1,7,1)*150.
mods_d1 = proportions_d1 * N_MSN
mods_d2 = proportions_d2 * N_MSN
raw_r = []
conv_r = []
mean_conv_r = []
std_conv_r = []
m_d1 = 1000
m_d2 = 0
inputs = []
i = 0
if not load:
tmp_rates = []
for e in range(n_exp):
seed = i
rates_SNR, rates_GPE = simulate_network_direct_indirect_onoff_vs_rate(
m_d1,
m_d1,
params_msn_d1,
params_msn_d2,
params_stn,
synapse_models, {
'SNR': 280,
'STN': 0,
'GPE': 20
},
sim_time=sim_time,
seed=seed,
threads=threads,
start_rec=start_rec,
model_params=model_params,
flag_bg=flag_bg)
tmp_rates.append(list(rates_SNR))
raw_r.append(numpy.array(tmp_rates))
i += 1
raw_r.append(numpy.array(tmp_rates))
inputs.append((m_d1, m_d2))
misc.pickle_save([raw_r, conv_r, mean_conv_r, std_conv_r, inputs],
save_at)
else:
raw_r, conv_r, mean_conv_r, std_conv_r, inputs = misc.pickle_load(
save_at)
conv_r = []
mean_conv_r = []
std_conv_r = []
conv_r, mean_conv_r, std_conv_r = conv_data(raw_r,
conv_r,
mean_conv_r,
std_conv_r,
bin=1,
kernel='triangle')
# pylab.plot(mean_conv_r[-1])
# pylab.plot(conv_r[-1][0])
# pylab.plot(conv_r[-1][1])
# pylab.plot(mean_conv_r[-1]-std_conv_r[-1],'--k')
# pylab.plot(mean_conv_r[-1]+std_conv_r[-1],'--k')
# pylab.show()
return raw_r, conv_r
def simulate_direct_vs_indirect(load,
N_MSN,
burst_rate,
save_at,
threads,
resolution=10,
flag_bg=False):
params_msn_d1 = {
'base_rates': [0.1],
'base_times': [1],
'mod_rates': [0.1, burst_rate, 0.1],
'mod_times': [1, 1000, 1000 + 500],
'n_mod': 0,
'bg_rate': 0
}
params_msn_d2 = {
'base_rates': [0.1],
'base_times': [1],
'mod_rates': [0.1, burst_rate, 0.1],
'mod_times': [1, 1000, 1000 + 500],
'n_mod': 0,
'focus': False,
'skip': 1,
'bg_rate': 0
}
params_stn = {
'rate': 250.,
'mod': False,
'mod_rate': 0.,
'mod_times': [1000., 1000. + 500.]
}
model_params = {
'conns': {
'MSN_D2_GPE': {
'lines': False
}
},
'neurons': {
'MSN_D1': {
'n': N_MSN
},
'MSN_D2': {
'n': N_MSN
}
}
}
synapse_models = ['MSN_SNR_gaba_p1', 'GPE_SNR_gaba_p']
if flag_bg: save_at = save_at + '_bg'
sim_time = 2000.
n_exp = 2
proportions_d1 = numpy.linspace(0.01, 0.15,
resolution) #arange(1,7,1)*150.
proportions_d2 = numpy.linspace(0.01, 0.15,
resolution) #arange(1,7,1)*150.
mods_d1 = proportions_d1 * N_MSN
mods_d2 = proportions_d2 * N_MSN
raw_r = []
conv_r = []
mean_conv_r = []
std_conv_r = []
inputs = []
i = 0
if not load:
for m_d1 in mods_d1:
for m_d2 in mods_d2:
tmp_rates = []
for e in range(n_exp):
seed = i
rates_SNR, rates_GPE = simulate_network_direct_indirect_onoff_vs_rate(
m_d1,
m_d1,
params_msn_d1,
params_msn_d2,
params_stn,
synapse_models, {
'SNR': 280,
'STN': 0,
'GPE': 20
},
sim_time=sim_time,
seed=seed,
threads=threads,
start_rec=500.,
model_params=model_params,
flag_bg=flag_bg)
tmp_rates.append(list(rates_SNR))
i += 1
raw_r.append(numpy.array(tmp_rates))
conv_r.append(
misc.convolve(raw_r[-1], 100, 'rectangle', single=False))
mean_conv_r.append(numpy.mean(conv_r[-1], axis=0))
std_conv_r.append(numpy.std(conv_r[-1], axis=0))
inputs.append((m_d1, m_d2))
misc.pickle_save(
[raw_r, conv_r, mean_conv_r, std_conv_r, inputs], save_at)
else:
raw_r, conv_r, mean_conv_r, std_conv_r, inputs = misc.pickle_load(
save_at)
data, data_split, slopes = get_data_and_slopes(inputs, mean_conv_r,
std_conv_r, resolution,
N_MSN)
return data, data_split, slopes, burst_rate, proportions_d1, proportions_d2
def simulate_indirect(load,
N_MSN,
save_at,
threads,
resolution=10,
max_prop=1.1,
flag_bg=False,
skip=1,
lines=False):
params_msn_d1 = {
'base_rates': [0.1],
'base_times': [1],
'mod_rates': [],
'mod_times': [],
'n_mod': 0,
'bg_rate': 0
}
params_msn_d2 = {
'base_rates': [0.1],
'base_times': [1],
'mod_rates': [],
'mod_times': [],
'n_mod': 0,
'focus': False,
'skip': skip,
'bg_rate': 0
}
params_stn = {
'rate': 250.,
'mod': False,
'mod_rate': 0.,
'mod_times': [1000., 1000. + 500.]
}
synapse_models = ['MSN_SNR_gaba_p1', 'GPE_SNR_gaba_p']
model_params = {
'misc': {
'N_MSN': N_MSN
},
'conns': {
'MSN_D2_GPE': {
'lines': lines
}
},
'neurons': {
'MSN_D1': {
'n': N_MSN
},
'MSN_D2': {
'n': N_MSN
}
}
}
if flag_bg: save_at = save_at + '_bg'
sim_time = 2000.
n_exp = 5
burst_rate = numpy.linspace(15, 50, resolution)
proportions = numpy.linspace(0.01, max_prop,
resolution) #arange(1,7,1)*150.
mods = proportions * N_MSN
raw_r, conv_r, mean_conv_r, std_conv_r, inputs = [], [], [], [], []
i = 0
if not load:
for m in mods:
for r in burst_rate:
tmp_rates = []
for e in range(n_exp):
seed = i
rates_SNR, rates_GPE = simulate_indirect_fun(
m,
r,
params_msn_d1,
params_msn_d2,
params_stn,
synapse_models, {
'SNR': 280,
'STN': 0,
'GPE': 20
},
sim_time=sim_time,
seed=seed,
threads=threads,
start_rec=500.,
model_params=model_params,
flag_bg=flag_bg,
max_mod=max(mods))
tmp_rates.append(list(rates_SNR))
i += 1
raw_r.append(numpy.array(tmp_rates))
conv_r.append(
misc.convolve(raw_r[-1], 100, 'rectangle', single=False))
mean_conv_r.append(numpy.mean(conv_r[-1], axis=0))
std_conv_r.append(numpy.std(conv_r[-1], axis=0))
inputs.append((m, r))
misc.pickle_save(
[raw_r, conv_r, mean_conv_r, std_conv_r, inputs], save_at)
else:
raw_r, conv_r, mean_conv_r, std_conv_r, inputs = misc.pickle_load(
save_at)
data, data_split, slopes = get_data_and_slopes(inputs, mean_conv_r,
std_conv_r, resolution,
N_MSN)
return data, data_split, slopes, burst_rate, proportions
def simulate_selection_vs_neurons(selection_intervals=[0.0, 500.0],
hz=20,
load=True):
global SNR_INJECTED_CURRENT
global NEURON_MODELS
global N_GPE
global N_MSN_BURST
global N_MSN
global GPE_BASE_RATE
global FILE_NAME
global OUTPUT_PATH
global SYNAPSE_MODELS_TESTED
global SEL_ONSET
#n_exp=100
n_exp = 2
if hz > 7:
n_max_sel = 60
if hz > 20:
n_max_sel = 30
else:
n_max_sel = 100
RATE_BASE = 0.1
RATE_SELE = hz
save_result_at = (OUTPUT_PATH + '/' + FILE_NAME +
'-simulate_selection_vs_neurons' + str(hz) + '-hz.pkl')
save_header_at = (OUTPUT_PATH + '/' + FILE_NAME +
'-simulate_selection_vs_neurons' + str(hz) +
'-hz_header')
burst_time = 500.
sim_time = burst_time + SEL_ONSET + 500.
EXPERIMENTS = range(n_exp)
MODEL_LIST = models()
my_nest.ResetKernel()
my_nest.MyLoadModels(MODEL_LIST, NEURON_MODELS)
my_nest.MyLoadModels(MODEL_LIST, SYNAPSE_MODELS_TESTED)
my_nest.MyLoadModels(MODEL_LIST, SYNAPSE_MODELS_BACKGROUND)
MSN_list = [] # MSN input for each experiment
for i_exp in EXPERIMENTS:
MSN = MyPoissonInput(n=N_MSN + n_max_sel, sd=True)
MSN_list.append(MSN)
GPE_list = [] # GPE input for each experiment
for i_exp in EXPERIMENTS:
GPE = MyPoissonInput(n=N_GPE, sd=True)
GPE_list.append(GPE)
SNR_list = [] # SNR groups for each synapse and number of selected MSN
SNR_list_experiments = []
for i_syn, syn in enumerate(SYNAPSE_MODELS_TESTED):
SNR = []
for i_sel in range(n_max_sel + 1): # Plus one to get no burst point
I_e = my_nest.GetDefaults(
NEURON_MODELS[0])['I_e'] + SNR_INJECTED_CURRENT
SNR.append(
MyGroup(NEURON_MODELS[0],
n=n_exp,
sd=True,
params={'I_e': I_e}))
SNR_list.append(SNR)
if not load:
for i_exp in EXPERIMENTS:
MSN = MSN_list[i_exp]
GPE = GPE_list[i_exp]
# Set spike times
# Base rate
for id in MSN[1:N_MSN]:
MSN.set_spike_times(id=id,
rates=[RATE_BASE],
times=[1],
t_stop=sim_time,
seed=int(numpy.random.random() * 10000.0))
# Selection
for id in MSN[N_MSN:N_MSN + n_max_sel]:
rates = [RATE_BASE, RATE_SELE, RATE_BASE]
times = [1, SEL_ONSET, burst_time + SEL_ONSET]
t_stop = sim_time
MSN.set_spike_times(id=id,
rates=rates,
times=times,
t_stop=t_stop,
seed=int(numpy.random.random() * 10000.0))
# Base rate GPE
for id in GPE[:]:
GPE.set_spike_times(id=id,
rates=[GPE_BASE_RATE],
times=[1],
t_stop=sim_time,
seed=int(numpy.random.random() * 10000.0))
# Connect
for i_syn, syn in enumerate(SYNAPSE_MODELS_TESTED):
# i_sel goes over 0,..., n_max_sel
for i_sel in range(0, n_max_sel + 1):
target = SNR_list[i_syn][i_sel][i_exp]
my_nest.ConvergentConnect(MSN[0:N_MSN - i_sel], [target],
model=syn)
my_nest.ConvergentConnect(MSN[N_MSN:N_MSN + i_sel],
[target],
model=syn)
my_nest.ConvergentConnect(
GPE[:], [target], model=SYNAPSE_MODELS_BACKGROUND[0])
my_nest.MySimulate(sim_time)
for SNR_sel in SNR_list:
for SNR in SNR_sel:
SNR.get_signal('s')
sel_interval_mean_rates = []
sel_interval_mean_rates_std = []
for i_interval, interval in enumerate(selection_intervals):
t1 = selection_intervals[i_interval][0]
t2 = selection_intervals[i_interval][1]
mean_rates = []
mean_rates_std = []
# Time until arrival of spikes in SNr
delay = my_nest.GetDefaults(SYNAPSE_MODELS_BACKGROUND[0])['delay']
for SNR_sel in SNR_list:
m_r = []
m_r_std = []
for SNR in SNR_sel:
m_r.append(SNR.signals['spikes'].mean_rate(
SEL_ONSET + t1 + delay, SEL_ONSET + t2 + delay))
m_r_std.append(SNR.signals['spikes'].mean_rate_std(
SEL_ONSET + t1 + delay, SEL_ONSET + t2 + delay))
mean_rates.append(m_r)
mean_rates_std.append(m_r_std)
mean_rates = numpy.array(mean_rates)
mean_rates_std = numpy.array(mean_rates_std)
sel_interval_mean_rates.append(mean_rates)
sel_interval_mean_rates_std.append(mean_rates_std)
nb_neurons = numpy.arange(0, n_max_sel + 1, 1)
s = '\n'
s = s + ' %s %5s %3s \n' % ('N MSNs:', str(N_MSN), '#')
s = s + ' %s %5s %3s \n' % ('N experiments:', str(n_exp), '#')
s = s + ' %s %5s %3s \n' % ('MSN base rate:', str(MSN_BASE_RATE), 'Hz')
s = s + ' %s %5s %3s \n' % ('MSN burst rate:', str(MSN_BURST_RATE),
'Hz')
s = s + ' %s %5s %3s \n' % ('GPe rate:', str(GPE_BASE_RATE), 'Hz')
s = s + ' %s %5s %3s \n' % ('Burst time:', str(burst_time), 'ms')
s = s + ' %s %5s %3s \n' % ('SNR_INJECTED_CURRENT:',
str(SNR_INJECTED_CURRENT), 'pA')
for i_interval, interval in enumerate(selection_intervals):
s = s + ' %s %5s %3s \n' % ('Sel interval ' + str(i_interval) +
':', str(selection_intervals), 'ms')
info_string = s
header = HEADER_SIMULATION_SETUP + s
misc.text_save(header, save_header_at)
misc.pickle_save([
nb_neurons, sel_interval_mean_rates, sel_interval_mean_rates_std,
info_string
], save_result_at)
elif load:
nb_neurons, sel_interval_mean_rates, sel_interval_mean_rates_std, info_string = misc.pickle_load(
save_result_at)
return nb_neurons, sel_interval_mean_rates, sel_interval_mean_rates_std, info_string
info_string = info_string + s
load_fast = True
save_result_at = OUTPUT_PATH + '/' + NEURON_MODELS[
0] + 'simulate_selection_vs_neurons_full_load_fast.pkl'
if not load_fast:
sel_interval_1 = [0.0, 500.0]
nbNeurons1, meanRates1, meanRatesStd1, s = simulate_selection_vs_neurons(
sel_interval_1, hz=20, load=True)
s = s + ' %s %5s %3s \n' % ('Sel thr:', str(SELECTION_THR), 'Hz')
s = s + ' \n%s %5s %3s \n' % ('Sel interval 1:', str(sel_interval_1),
'pA')
info_string = info_string + s
misc.pickle_save([nbNeurons1, meanRates1, meanRatesStd1, info_string],
save_result_at)
else:
print 'hej'
#nbNeurons1, meanRates1, meanRatesStd1, info_string = misc.pickle_load(save_result_at)
# DISPLAY
plot_settings.set_mode(pylab,
mode='by_fontsize',
w=1100.0,
h=450.0,
fontsize=12)
font_size_text = 8
fig = pylab.figure(facecolor='w')
ax_list = []
ax_list.append(MyAxes(fig, [.075, .37, .135, .26])) # text box
load=False
MSN_list_sel2, GPE_list_sel2, SNR_list_sel2, s2=simulate_example(MSN_hz=20,
GPE_hz=100,
load=load,
n_gpe_sel=N_GPE_BURST,
sel_time_GPE=GPE_BURST_TIME)
MSN_list_inh, GPE_list_inh, SNR_list_inh, s3=simulate_example(MSN_hz=20,
GPE_hz=0,
load=load,
n_gpe_sel=N_GPE_PAUSE,
sel_time_GPE=GPE_PAUSE_TIME)
info_string=info_string+s2+s3
misc.pickle_save([SNR_list_sel2,SNR_list_inh, info_string] , save_at)
else:
SNR_list_sel2,SNR_list_inh, info_string=misc.pickle_load(save_at)
# DISPLAY
plot_settings.set_mode(pylab, mode='by_fontsize', w = 1100.0, h = 450.0, fontsize=12)
font_size_text = 8
fig = pylab.figure( facecolor = 'w' )
ax_list = []
ax_list.append( MyAxes(fig, [ .075, .37, .135, .26 ] ) ) # text box
ax_list.append( MyAxes(fig, [ .26, .6, .165*2.312, .34 ] ) ) #
ax_list.append( MyAxes(fig, [ .26, .1, .165*2.312, .34 ] ) ) #
#ax_list.append( MyAxes(fig, [ .8, .6, .165, .34 ] ) ) #
#ax_list.append( MyAxes(fig, [ .26, .1, .165, .34 ] ) ) #
def simulate_example(MSN_hz=20, GPE_hz=0, load=True, n_gpe_sel=3, sel_time_GPE=500):
global GPE_BASE_RATE
global STN_BASE_RATE
global MSN_BASE_RATE
global MSN_BURST_TIME
global NEURON_MODELS
global N_GPE
global N_STN
global N_MSN
global N_MSN_BURST
global SNAME
global SPATH
global SYNAPSE_MODELS
global SEL_ONSET
global SNR_INJECTED_CURRENT
n_exp = 200
msn_rate_sel = MSN_hz # Selection rate
gpe_sel_rate = GPE_hz # Selection rate
sel_time_MSN = MSN_BURST_TIME
sim_time = sel_time_MSN+SEL_ONSET+500.
EXPERIMENTS=range(n_exp)
MODEL_LIST=models()
my_nest.ResetKernel()
my_nest.MyLoadModels( MODEL_LIST, NEURON_MODELS )
my_nest.MyLoadModels( MODEL_LIST, SYNAPSE_MODELS)
my_nest.MyLoadModels( MODEL_LIST, SYNAPSE_MODELS_BACKGROUND)
MSN_list=[] # MSN input for each experiment
for i_exp in EXPERIMENTS:
MSN = MyPoissonInput( n=N_MSN+N_MSN_BURST, sd=True)
MSN_list.append(MSN)
GPE_list=[] # GPE input for each experiment
for i_exp in EXPERIMENTS:
GPE = MyPoissonInput( n=N_GPE+n_gpe_sel, sd=True)
GPE_list.append(GPE)
STN_list=[] # GPE input for each experiment
for i_exp in EXPERIMENTS:
STN = MyPoissonInput( n=N_STN, sd=True)
STN_list.append(GPE)
SNR_list=[] # SNR groups for each synapse
for i, SNR_i_c in enumerate(SNR_INJECTED_CURRENT):
I_e=my_nest.GetDefaults(NEURON_MODELS[0])['I_e']+SNR_i_c
SNR = MyGroup( NEURON_MODELS[0], n=n_exp, params={'I_e':I_e},
sd=True, mm=False,
mm_dt=.1, record_from=[''])
SNR_list.append(SNR)
if not load:
for i_exp in EXPERIMENTS:
# MSN
MSN = MSN_list[i_exp]
# Set spike times
# Base rate
for id in MSN[0:N_MSN]:
MSN.set_spike_times(id=id, rates=[MSN_BASE_RATE], times=[1],
t_stop=sim_time,
seed=int(numpy.random.random()*10000.0))
# Selection MSN
for id in MSN[N_MSN:N_MSN+N_MSN_BURST]:
rates = [MSN_BASE_RATE, msn_rate_sel, MSN_BASE_RATE]
times = [1, SEL_ONSET, sel_time_MSN + SEL_ONSET]
t_stop = sim_time
MSN.set_spike_times(id=id, rates=rates, times=times,
t_stop=t_stop,
seed=int(numpy.random.random()*10000.0))
# GPE
GPE = GPE_list[i_exp]
# Set spike times
# Base rate
for id in GPE[:]:
GPE.set_spike_times(id=id, rates=[GPE_BASE_RATE], times=[1],
t_stop=sim_time,
seed=int(numpy.random.random()*10000.0))
# Selection GPE
for id in GPE[N_GPE:N_GPE+n_gpe_sel]:
rates = [GPE_BASE_RATE, gpe_sel_rate, GPE_BASE_RATE]
# If GPe excited smaller selection time
times = [1, SEL_ONSET, sel_time_GPE + SEL_ONSET]
t_stop = sim_time
GPE.set_spike_times(id=id, rates=rates, times=times,
t_stop=t_stop, seed=int(numpy.random.random()*100000.0))
# Base rate STN
for id in STN[:]:
STN.set_spike_times(id=id, rates=[STN_BASE_RATE], times=[1],
t_stop=sim_time,
seed=int(numpy.random.random()*10000.0))
idx_MSN_s=range(0,N_MSN-N_MSN_BURST)
idx_MSN_s.extend(range(N_MSN,N_MSN+N_MSN_BURST))
idx_GPE_s=range(0,N_GPE-n_gpe_sel)
idx_GPE_s.extend(range(N_GPE,N_GPE+n_gpe_sel))
# Connect with MSN burst
target=SNR_list[0][i_exp]
my_nest.ConvergentConnect(MSN[idx_MSN_s], [target], model=SYNAPSE_MODELS[0])
my_nest.ConvergentConnect(GPE[0:N_GPE], [target], model=SYNAPSE_MODELS[1])
my_nest.ConvergentConnect(STN[:], [target], model=SYNAPSE_MODELS_BACKGROUND[0])
# With GPe pause
target=SNR_list[1][i_exp]
my_nest.ConvergentConnect(MSN[0:N_MSN], [target], model=SYNAPSE_MODELS[0])
my_nest.ConvergentConnect(GPE[idx_GPE_s], [target], model=SYNAPSE_MODELS[1])
my_nest.ConvergentConnect(STN[:], [target], model=SYNAPSE_MODELS_BACKGROUND[0])
# With MSN burst and GPe pause
target=SNR_list[2][i_exp]
my_nest.ConvergentConnect(MSN[idx_MSN_s], [target], model=SYNAPSE_MODELS[0])
my_nest.ConvergentConnect(GPE[idx_GPE_s], [target], model=SYNAPSE_MODELS[1])
my_nest.ConvergentConnect(STN[:], [target], model=SYNAPSE_MODELS_BACKGROUND[0])
my_nest.MySimulate( sim_time )
for MSN in MSN_list:
MSN.get_signal( 's' )
for GPE in GPE_list:
GPE.get_signal( 's' )
for SNR in SNR_list:
SNR.get_signal( 's' )
misc.pickle_save([MSN_list, GPE_list,SNR_list] , save_at)
if load:
MSN_list, GPE_list, SNR_list=misc.pickle_load(save_at)
pre_dyn_MSN=str(SNR_list[0].signals['spikes'].mean_rate(SEL_ONSET-500,
SEL_ONSET))
pre_dyn_GPE=str(SNR_list[1].signals['spikes'].mean_rate(SEL_ONSET-500,
SEL_ONSET))
s='\n'
s=s+'Example:\n'
s = s + ' %s %5s %3s \n' % ( 'N experiments:', str ( len(EXPERIMENTS) ), '#' )
s = s + ' %s %5s %3s \n' % ( 'N MSN:', str ( N_MSN ), '#' )
s = s + ' %s %5s %3s \n' % ( 'N GPE:', str ( N_GPE ), '#' )
s='\n'
s = s + ' %s %5s %3s \n' % ( 'Base rate MSN:', str ( MSN_BASE_RATE),'spikes/s' )
s = s + ' %s %5s %3s \n' % ( 'Sel rate MSN:', str ( msn_rate_sel ), 'spikes/s' )
s = s + ' %s %5s %3s \n' % ( 'Sel time MSN:', str ( sel_time_MSN ), 'ms' )
s='\n'
s = s + ' %s %5s %3s \n' % ( 'Base rate GPe:', str ( GPE_BASE_RATE),'spikes/s' )
s = s + ' %s %5s %3s \n' % ( 'Sel rate GPe:', str ( gpe_sel_rate ), 'spikes/s' )
s = s + ' %s %5s %3s \n' % ( 'Sel time GPe:', str ( sel_time_GPE ), 'ms' )
s = s + ' %s %5s %3s \n' % ( 'Pre sel rate Dyn MSN:', pre_dyn_MSN[0:4], 'spikes/s' )
s = s + ' %s %5s %3s \n' % ( 'Pre sel rate Dyn GPe:', pre_dyn_GPE[0:4], 'spikes/s' )
return MSN_list, GPE_list, SNR_list, s
info_string=s
return MSN_hzs, GPE_hzs, data, info_string
def simulate_example(load=True):
global SNR_INJECTED_CURRENT
global NEURON_MODELS
global N_GPE
global N_MSN_BURST
global N_MSN
global GPE_BASE_RATE
global FILE_NAME
global OUTPUT_PATH
global SYNAPSE_MODELS_TESTED
global SEL_ONSET
#n_exp =200 # number of experiments
n_exp = 20 # number of experiments
# Path were raw data is saved. For example the spike trains.
save_result_at = OUTPUT_PATH + '/' + FILE_NAME + '-simulate_example.pkl'
save_header_at = OUTPUT_PATH + '/' + FILE_NAME + '-simulate_example_header'
burst_time = 500.
sim_time = burst_time + SEL_ONSET + 1000.
MODEL_LIST = models()
my_nest.ResetKernel()
my_nest.MyLoadModels(MODEL_LIST, NEURON_MODELS)
my_nest.MyLoadModels(MODEL_LIST, SYNAPSE_MODELS_TESTED)
my_nest.MyLoadModels(MODEL_LIST, SYNAPSE_MODELS_BACKGROUND)
SNR_list = [] # List with SNR groups for synapse.
if not load:
MSN_base = MyPoissonInput(n=N_MSN_BASE * n_exp, sd=True)
MSN_burst = MyPoissonInput(n=N_MSN_BURST * n_exp, sd=True)
GPE = MyPoissonInput(n=N_GPE * n_exp, sd=True)
# Set spike times MSN and GPe
# Non bursting MSNs
for id in MSN_base[:]:
seed = numpy.random.random_integers(0, 1000000.0)
MSN_base.set_spike_times(id=id,
rates=[MSN_BASE_RATE],
times=[1],
t_stop=sim_time,
seed=seed)
# Background GPe
for id in GPE[:]:
seed = numpy.random.random_integers(0, 1000000.0)
GPE.set_spike_times(id=id,
rates=[GPE_BASE_RATE],
times=[1],
t_stop=sim_time,
seed=seed)
# Bursting MSNs
for id in MSN_burst[:]:
rates = [MSN_BASE_RATE, MSN_BURST_RATE, MSN_BASE_RATE]
times = [1, SEL_ONSET, burst_time + SEL_ONSET]
t_stop = sim_time
seed = numpy.random.random_integers(0, 1000000.0)
MSN_burst.set_spike_times(id=id,
rates=rates,
times=times,
t_stop=t_stop,
seed=seed)
for i_syn in range(len(SYNAPSE_MODELS_TESTED)):
params = []
I_e = my_nest.GetDefaults(
NEURON_MODELS[0])['I_e'] + SNR_INJECTED_CURRENT
for i in range(n_exp):
#params.append({'I_e':numpy.random.normal(I_e,
# 0.1*I_e)})
params.append({'I_e': I_e})
#{'I_e':SNR_INJECTED_CURRENT}
SNR = MyGroup(NEURON_MODELS[0],
n=n_exp,
sd=True,
params=params,
mm_dt=.1,
record_from=[''])
SNR_list.append(SNR)
# Connect, experiment specific
sources_MSN_SNR_base = numpy.arange(0, n_exp * N_MSN_BASE)
sources_MSN_SNR_burst = numpy.arange(0, n_exp * N_MSN_BURST)
targets_MSN_SNR_base = numpy.mgrid[0:n_exp, 0:N_MSN_BASE][0].reshape(
1, N_MSN_BASE * n_exp)[0]
targets_MSN_SNR_burst = numpy.mgrid[0:n_exp, 0:N_MSN_BURST][0].reshape(
1, N_MSN_BURST * n_exp)[0]
sources_GPE_SNR = numpy.arange(0, n_exp * N_GPE)
targets_GPE_SNR = numpy.mgrid[0:n_exp,
0:N_GPE][0].reshape(1, N_GPE * n_exp)[0]
for i_syn, syn in enumerate(SYNAPSE_MODELS_TESTED):
syn = SYNAPSE_MODELS_TESTED[i_syn]
SNR = SNR_list[i_syn]
my_nest.Connect(MSN_base[sources_MSN_SNR_base],
SNR[targets_MSN_SNR_base],
model=syn)
my_nest.Connect(MSN_burst[sources_MSN_SNR_burst],
SNR[targets_MSN_SNR_burst],
model=syn)
my_nest.Connect(GPE[sources_GPE_SNR],
SNR[targets_GPE_SNR],
model=SYNAPSE_MODELS_BACKGROUND[0])
my_nest.MySimulate(sim_time)
MSN_base.get_signal('s', start=0, stop=sim_time)
MSN_burst.get_signal('s', start=0, stop=sim_time)
for SNR in SNR_list:
SNR.get_signal('s', start=0, stop=sim_time)
# Get firing rates of MSNs
MSN_firing_rates = []
MSN_all = copy.deepcopy(MSN_base)
MSN_all.merge(MSN_burst)
time_bin = 20.
groups = [MSN_base, MSN_burst, MSN_all]
for group in groups:
timeAxis, firingRates = group.signals['spikes'].my_firing_rate(
bin=time_bin, display=False)
MSN_firing_rates.append([timeAxis, firingRates])
# Pick out spikes for burst, base and all to use in scatter plot
MSN_spikes_and_ids = []
g1 = MSN_burst.slice(MSN_burst[0:N_MSN_BURST])
g2 = MSN_base.slice(MSN_base[0:N_MSN_BASE])
ids_MSN_burst = range(450, 450 + N_MSN_BURST)
ids_MSN_base = [id for id in range(N_MSN) if id not in IDS_MSN_BURST]
# Rename ids for plotting purpose
g1_dict = dict([[id1, id2] for id1, id2 in zip(g1.ids, ids_MSN_burst)])
g2_dict = dict([[id1, id2] for id1, id2 in zip(g2.ids, ids_MSN_base)])
groups = [g1, g2]
dics = [g1_dict, g2_dict]
for group, dic in zip(groups, dics):
raw_data = group.signals['spikes'].raw_data()
for i in range(raw_data.shape[0]):
raw_data[i, 1] = dic[raw_data[i, 1]]
MSN_spikes_and_ids.append(raw_data)
#times, binned_data=MSN_base.signals['spikes'].binned_raw_data(0, sim_time, res=1, clip=0)
#filtered_binned_data=misc.time_resolved_rate(binned_data, 100, kernel_type='triangle', res=1)
pre_ref_1 = str(SNR_list[0].signals['spikes'].mean_rate(
SEL_ONSET - 500, SEL_ONSET))
pre_ref_2 = str(SNR_list[1].signals['spikes'].mean_rate(
SEL_ONSET - 500, SEL_ONSET))
pre_dyn = str(SNR_list[2].signals['spikes'].mean_rate(
SEL_ONSET - 500, SEL_ONSET))
s = '\n'
s = s + 'Simulate example:\n'
s = s + '%s %5s %3s \n' % ('N experiments:', str(n_exp), '#')
s = s + '%s %5s %3s \n' % ('Bin size MSN hz:', str(time_bin), 'ms')
s = s + '%s %5s %3s \n' % ('MSN base rate:', str(MSN_BASE_RATE), 'Hz')
s = s + '%s %5s %3s \n' % ('MSN burst rate:', str(MSN_BURST_RATE),
'Hz')
s = s + '%s %5s %3s \n' % ('GPe rate:', str(GPE_BASE_RATE), 'Hz')
s = s + '%s %5s %3s \n' % ('Burst time:', str(burst_time), 'ms')
s = s + '%s %5s %3s \n' % ('Pre sel rate Ref:', pre_ref_1[0:4], 'Hz')
s = s + '%s %5s %3s \n' % ('Pre sel rate Ref:', pre_ref_2[0:4], 'Hz')
s = s + '%s %5s %3s \n' % ('Pre sel rate Dyn:', pre_dyn[0:4], 'Hz')
header = HEADER_SIMULATION_SETUP + s
misc.pickle_save([MSN_firing_rates, MSN_spikes_and_ids, SNR_list, s],
save_result_at)
misc.text_save(header, save_header_at)
else:
MSN_firing_rates, MSN_spikes_and_ids, SNR_list, s = misc.pickle_load(
save_result_at)
return MSN_firing_rates, MSN_spikes_and_ids, SNR_list, s
def simulate_example(hz_1=0., hz_2=100., load=True):
global I_E
global NEURON_MODELS
global N_GPE
global N_SEL
global N_MSN
global N_STN
global MSN_RATE_BASE
global STN_RATE_BASE
global SNAME
global SPATH
global SYNAPSE_MODELS
global SEL_ONSET
N_EXP = 200
RATE_BASE = 25 # Base rate
RATE_SELE_1 = hz_1
RATE_SELE_2 = hz_2 # Selection rate
SAVE_AT = SPATH + '/' + NEURON_MODELS[0] + '-example.pkl'
SEL_TIME_1 = 500.
SEL_TIME_2 = 200.
sim_time = SEL_TIME_1 + SEL_TIME_2 + SEL_ONSET + 500.
SNAME_NB = hz_1 + hz_2 + 1000
EXPERIMENTS = range(N_EXP)
MODEL_LIST = models()
my_nest.ResetKernel()
my_nest.MyLoadModels(MODEL_LIST, NEURON_MODELS)
my_nest.MyLoadModels(MODEL_LIST, SYNAPSE_MODELS_TESTED)
my_nest.MyLoadModels(MODEL_LIST, SYNAPSE_MODELS_BACKGROUND)
GPE_list = [] # GPE input for each experiment
for i_exp in EXPERIMENTS:
GPE = MyPoissonInput(n=N_GPE, sd=True)
GPE_list.append(GPE)
MSN_list = [] # MSN input for each experiment
for i_exp in EXPERIMENTS:
MSN = MyPoissonInput(n=N_MSN, sd=True)
MSN_list.append(MSN)
STN_list = [] # MSN input for each experiment
for i_exp in EXPERIMENTS:
STN = MyPoissonInput(n=N_STN, sd=True)
STN_list.append(STN)
SNR_list = [] # SNR groups for each synapse
for i_syn in SYNAPSE_MODELS_TESTED:
I_e = my_nest.GetDefaults(NEURON_MODELS[0])['I_e'] + I_E
SNR = MyGroup(NEURON_MODELS[0],
n=N_EXP,
params={'I_e': I_e},
sd=True,
mm=False,
mm_dt=.1,
record_from=[''])
SNR_list.append(SNR)
if not load:
for i_exp in EXPERIMENTS:
GPE = GPE_list[i_exp]
MSN = MSN_list[i_exp]
STN = STN_list[i_exp]
# Set spike times
# Base rate MSN
for id in MSN[:]:
MSN.set_spike_times(id=id,
rates=[MSN_RATE_BASE],
times=[1],
t_stop=sim_time,
seed=int(numpy.random.random() * 10000.0))
# Base rate
for id in GPE[0:N_GPE - N_SEL]:
GPE.set_spike_times(id=id,
rates=[RATE_BASE],
times=[1],
t_stop=sim_time,
seed=int(numpy.random.random() * 10000.0))
# Base rate STN
for id in STN[:]:
STN.set_spike_times(id=id,
rates=[STN_RATE_BASE],
times=[1],
t_stop=sim_time,
seed=int(numpy.random.random() * 10000.0))
# Selection
for id in GPE[N_GPE - N_SEL:N_GPE]:
rates = [RATE_BASE, RATE_SELE_1, RATE_SELE_2, RATE_BASE]
times = [
1, SEL_ONSET, SEL_ONSET + SEL_TIME_1,
SEL_ONSET + SEL_TIME_1 + SEL_TIME_2
]
t_stop = sim_time
GPE.set_spike_times(id=id,
rates=rates,
times=times,
t_stop=t_stop,
seed=int(numpy.random.random() * 10000.0))
# Connect
for i, syn in enumerate(SYNAPSE_MODELS_TESTED):
target = SNR_list[i][i_exp]
my_nest.ConvergentConnect(GPE[:], [target], model=syn)
my_nest.ConvergentConnect(MSN[:], [target],
model=SYNAPSE_MODELS_BACKGROUND[0])
my_nest.ConvergentConnect(STN[:], [target],
model=SYNAPSE_MODELS_BACKGROUND[1])
my_nest.MySimulate(sim_time)
for GPE in GPE_list:
GPE.get_signal('s')
for SNR in SNR_list:
SNR.get_signal('s')
misc.pickle_save([GPE_list, SNR_list], SAVE_AT)
elif load:
GPE_list, SNR_list = misc.pickle_load(SAVE_AT)
pre_ref = str(SNR_list[0].signals['spikes'].mean_rate(
SEL_ONSET - 500, SEL_ONSET))
pre_dyn = str(SNR_list[1].signals['spikes'].mean_rate(
SEL_ONSET - 500, SEL_ONSET))
statusSynapes = []
for syn in SYNAPSE_MODELS_TESTED:
statusSynapes.append(my_nest.GetDefaults(syn))
s = '\n'
s = s + 'Example:\n'
s = s + ' %s %5s %3s \n' % ('N experiments:', str(len(EXPERIMENTS)), '#')
s = s + ' %s %5s %3s \n' % ('N GPEs:', str(N_GPE), '#')
s = s + ' %s %5s %3s \n' % ('Base rate:', str(RATE_BASE), 'spikes/s')
s = s + ' %s %5s %3s \n' % ('Selection rate:', str(RATE_SELE_1),
'spikes/s')
s = s + ' %s %5s %3s \n' % ('Selection time:', str(SEL_TIME_1), 'ms')
s = s + ' %s %5s %3s \n' % ('Selection rate:', str(RATE_SELE_2),
'spikes/s')
s = s + ' %s %5s %3s \n' % ('Selection time:', str(SEL_TIME_2), 'ms')
s = s + ' %s %5s %3s \n' % ('Pre sel rate Ref:', pre_ref[0:4], 'spikes/s')
s = s + ' %s %5s %3s \n' % ('Pre sel rate Dyn:', pre_dyn[0:4], 'spikes/s')
for ss in statusSynapes:
s = s + '\n'
s = s + ' %s %10s\n' % ('Synapse', ss['synapsemodel'])
s = s + ' %s %5s %3s\n' % ('Weight', str(round(ss['weight'], 1)), 'nS')
return GPE_list, SNR_list, s
def simulate_selection_vs_neurons_full(selRateInterval,
load_pickle=True,
load_raw=True):
global N_MSN
save_result_at = OUTPUT_PATH + '/' + FILE_NAME + '-simulate_selection_vs_neurons_full.pkl'
save_result_at = OUTPUT_PATH + '/' + FILE_NAME + '-simulate_selection_vs_neurons_full_header'
selection_intervals = [[0, 200], [300, 500]]
#Range 1
#hzs=range(5,8)
# Range 2
#hzs=range(8,61,1) # MPI can cope with jump 7->8 when n max selected decreases
# Range
hzs = range(5, 49, 1)
#hzs=[8,20]
if not load_pickle:
data = {}
for syn in SYNAPSE_MODELS_TESTED:
data[syn] = {}
data[syn]['rates'] = [[] for k in range(len(selection_intervals))]
data[syn]['selMat'] = [[] for k in range(len(selection_intervals))]
data[syn]['thrVec'] = [[] for k in range(len(selection_intervals))]
for hz in hzs:
n, r, r_std, s = simulate_selection_vs_neurons(
selection_intervals, hz, load_raw)
print hz, 'hz finished'
for i_syn, syn in enumerate(SYNAPSE_MODELS_TESTED):
for k in range(len(selection_intervals)):
data[syn]['rates'][k].append(r[k][i_syn, :])
# Create matricies
# Adjust rates if they are different length, add zeros at end of
# short vectors. OBS only true if last rate is zero in vectorn that
# is elonged.
for k in range(len(selection_intervals)):
rates = data[syn]['rates'][k]
maxLen = 0
for r in rates:
if len(r) > maxLen:
maxLen = len(r)
for i_r, r in enumerate(rates):
rates[i_r] = numpy.append(
r, numpy.zeros((1, maxLen - len(r))))
selMat = rates
thrVec = []
for i in range(rates.shape[0]):
p = True
for j in range(rates.shape[1]):
if SELECTION_THR < r[i, j]:
selMat[i, j] = 3
elif (SELECTION_THR >= rates[i, j]) and (
SELECTION_THR < rates[i, j - 1]) and p:
selMat[i, j] = 2
thrVec.append(j + 1) # Neurons for threshold
p = False
else:
selMat[i, j] = 1
if p:
thrVec[j].append(100)
data[syn]['selMat'][k] = selMat
data[syn]['thrVec'][k] = numpy.array(thrVec)
if not mpiRun:
header = HEADER_SIMULATION_SETUP + s
misc.text_save(header, save_header_at)
misc.pickle_save(data, save_result_at)
elif load_pickle:
data = misc.pickle_load(save_result_at)
s = '\n'
info_string = s
return hzs, data, info_string
synapse_models = ['MSN_SNR_gaba_p1', 'GPE_SNR_gaba_p']
save_result_at = OUTPUT_PATH + '/simulate_target.plk'
p_weights = numpy.ones(17)
n_exp = 5
if 0:
r_target = []
for i in range(n_exp):
seed = i
r_SNR, r_GPE, r_STN = simulate(params_msn_d1, params_msn_d2,
params_stn, synapse_models, sim_time,
seed, {}, threads, start_rec,
model_params, p_weights)
r_target.append(r_SNR)
misc.pickle_save([r_target, r_GPE, r_STN], save_result_at)
else:
r_target, r_GPE, r_STN = misc.pickle_load(save_result_at)
r_target = numpy.array(r_target)
mr_target = numpy.mean(r_target, axis=0)
# Will find the value that GPE_SNR_ref synapse should be multiplied with
# so that it have the same effect in the network as the plastic at base-level
# network activty
x0 = 1.0
save_at = OUTPUT_PATH + '/simulate_network_fmin' + str(n_exp) + '.plk'
x, e = fmin(0, save_at, x0, n_exp, mr_target, params_msn_d1, params_msn_d2,
params_stn, sim_time, {}, threads, start_rec, model_params,
p_weights)
print x, e, r_target, r_GPE, r_STN