def write_spiketrains(self):
start_writing = time.time()
for layer in self.n.pops :
for pop in self.n.pops[layer] :
io = PyNNTextIO(filename=system_params['output_path'] \
+ "/spikes_" + layer \
+ '_' + pop \
+ '_' + str(sim.rank()) \
+ ".txt")
spikes = self.n.pops[layer][pop].get_data('spikes',
gather=False)
for segment in spikes.segments :
io.write_segment(segment)
if record_v :
io = PyNNTextIO(filename=system_params['output_path'] \
+ "/vm_" + layer \
+ '_' + pop \
+ '_' + str(sim.rank()) \
+ ".txt")
vm = self.n.pops[layer][pop].get_data('v',
gather=False)
for segment in vm.segments :
try :
io.write_segment(segment)
except AssertionError :
pass
end_writing = time.time()
print("Writing data took %g s" % (end_writing - start_writing,))
print('Creating the network took %g s' % (end_netw - start_netw,))
# simulate
if sim.rank() == 0 :
print("Simulating...")
start_sim = time.time()
t = sim.run(simulator_params[simulator]['sim_duration'])
end_sim = time.time()
if sim.rank() == 0 :
print('Simulation took %g s' % (end_sim - start_sim,))
start_writing = time.time()
for layer in n.pops :
for pop in n.pops[layer] :
io = PyNNTextIO(filename=system_params['output_path'] \
+ "/spikes_" + layer + '_' + pop + '_' + str(sim.rank()) + ".txt")
spikes = n.pops[layer][pop].get_data('spikes', gather=False)
for segment in spikes.segments :
io.write_segment(segment)
if record_v :
io = PyNNTextIO(filename=system_params['output_path'] \
+ "/vm_" + layer + '_' + pop + '_' + str(sim.rank()) + ".txt")
vm = n.pops[layer][pop].get_data('v', gather=False)
for segment in vm.segments :
try :
io.write_segment(segment)
except AssertionError :
pass
end_writing = time.time()
def runBrunelNetwork(g=5.,
eta=2.,
dt = 0.1,
simtime = 1000.0,
delay = 1.5,
epsilon = 0.1,
order = 2500,
N_rec = 50,
N_rec_v = 2,
save=False,
simulator_name='nest',
jnml_simulator=None,
extra = {}):
exec("from pyNN.%s import *" % simulator_name) in globals()
timer = Timer()
# === Define parameters ========================================================
downscale = 1 # scale number of neurons down by this factor
# scale synaptic weights up by this factor to
# obtain similar dynamics independent of size
order = order # determines size of network:
# 4*order excitatory neurons
# 1*order inhibitory neurons
Nrec = N_rec # number of neurons to record from, per population
epsilon = epsilon # connectivity: proportion of neurons each neuron projects to
# Parameters determining model dynamics, cf Brunel (2000), Figs 7, 8 and Table 1
# here: Case C, asynchronous irregular firing, ~35 Hz
eta = eta # rel rate of external input
g = g # rel strength of inhibitory synapses
J = 0.1 # synaptic weight [mV]
delay = delay # synaptic delay, all connections [ms]
# single neuron parameters
tauMem = 20.0 # neuron membrane time constant [ms]
tauSyn = 0.1 # synaptic time constant [ms]
tauRef = 2.0 # refractory time [ms]
U0 = 0.0 # resting potential [mV]
theta = 20.0 # threshold
# simulation-related parameters
simtime = simtime # simulation time [ms]
dt = dt # simulation step length [ms]
# seed for random generator used when building connections
connectseed = 12345789
use_RandomArray = True # use Python rng rather than NEST rng
# seed for random generator(s) used during simulation
kernelseed = 43210987
# === Calculate derived parameters =============================================
# scaling: compute effective order and synaptic strength
order_eff = int(float(order)/downscale)
J_eff = J*downscale
# compute neuron numbers
NE = int(4*order_eff) # number of excitatory neurons
NI = int(1*order_eff) # number of inhibitory neurons
N = NI + NE # total number of neurons
# compute synapse numbers
CE = int(epsilon*NE) # number of excitatory synapses on neuron
CI = int(epsilon*NI) # number of inhibitory synapses on neuron
C = CE + CI # total number of internal synapses per n.
Cext = CE # number of external synapses on neuron
# synaptic weights, scaled for alpha functions, such that
# for constant membrane potential, charge J would be deposited
fudge = 0.00041363506632638 # ensures dV = J at V=0
# excitatory weight: JE = J_eff / tauSyn * fudge
JE = (J_eff/tauSyn)*fudge
# inhibitory weight: JI = - g * JE
JI = -g*JE
# threshold, external, and Poisson generator rates:
nu_thresh = theta/(J_eff*CE*tauMem)
nu_ext = eta*nu_thresh # external rate per synapse
p_rate = 1000*nu_ext*Cext # external input rate per neuron (Hz)
# number of synapses---just so we know
Nsyn = (C+1)*N + 2*Nrec # number of neurons * (internal synapses + 1 synapse from PoissonGenerator) + 2synapses" to spike detectors
print('Created Brunel network with parameters')
for p in ['NE','NI','downscale','order','N_rec','epsilon','eta','g','J','delay','tauMem','tauSyn','tauRef','U0','theta','simtime','dt', \
'order_eff','J_eff','N','CE','CI','C','Cext','fudge','JE','JI','nu_thresh','nu_ext','p_rate','Nsyn']:
print(' %s%s= %s'%(p, ' '*(12-len(p)), eval(p)))
# put cell parameters into a dict
cell_params = {'tau_m' : tauMem,
'tau_syn_E' : tauSyn,
'tau_syn_I' : tauSyn,
'tau_refrac' : tauRef,
'v_rest' : U0,
'v_reset' : U0,
'v_thresh' : theta,
'cm' : 0.001} # (nF)
# === Build the network ========================================================
# clear all existing network elements and set resolution and limits on delays.
# For NEST, limits must be set BEFORE connecting any elements
#extra = {'threads' : 2}
rank = setup(timestep=dt, max_delay=delay, **extra)
print("rank =", rank)
np = num_processes()
print("np =", np)
import socket
host_name = socket.gethostname()
print("Host #%d is on %s" % (rank+1, host_name))
if 'threads' in extra:
print("%d Initialising the simulator with %d threads..." % (rank, extra['threads']))
else:
print("%d Initialising the simulator with single thread..." % rank)
# Small function to display information only on node 1
def nprint(s):
if rank == 0:
print(s)
timer.start() # start timer on construction
print("%d Setting up random number generator" % rank)
rng = NumpyRNG(kernelseed, parallel_safe=True)
print("%d Creating excitatory population with %d neurons." % (rank, NE))
celltype = IF_curr_alpha(**cell_params)
celltype.default_initial_values['v'] = U0 # Setting default init v, useful for NML2 export
E_net = Population(NE, celltype, label="E_net")
print("%d Creating inhibitory population with %d neurons." % (rank, NI))
I_net = Population(NI, celltype, label="I_net")
print("%d Initialising membrane potential to random values between %g mV and %g mV." % (rank, U0, theta))
uniformDistr = RandomDistribution('uniform', low=U0, high=theta, rng=rng)
E_net.initialize(v=uniformDistr)
I_net.initialize(v=uniformDistr)
print("%d Creating excitatory Poisson generator with rate %g spikes/s." % (rank, p_rate))
source_type = SpikeSourcePoisson(rate=p_rate)
expoisson = Population(NE, source_type, label="expoisson")
print("%d Creating inhibitory Poisson generator with the same rate." % rank)
inpoisson = Population(NI, source_type, label="inpoisson")
# Record spikes
print("%d Setting up recording in excitatory population." % rank)
E_net.record('spikes')
if N_rec_v>0:
E_net[0:min(NE,N_rec_v)].record('v')
print("%d Setting up recording in inhibitory population." % rank)
I_net.record('spikes')
if N_rec_v>0:
I_net[0:min(NI,N_rec_v)].record('v')
progress_bar = ProgressBar(width=20)
connector = FixedProbabilityConnector(epsilon, rng=rng, callback=progress_bar)
E_syn = StaticSynapse(weight=JE, delay=delay)
I_syn = StaticSynapse(weight=JI, delay=delay)
ext_Connector = OneToOneConnector(callback=progress_bar)
ext_syn = StaticSynapse(weight=JE, delay=dt)
print("%d Connecting excitatory population with connection probability %g, weight %g nA and delay %g ms." % (rank, epsilon, JE, delay))
E_to_E = Projection(E_net, E_net, connector, E_syn, receptor_type="excitatory")
print("E --> E\t\t", len(E_to_E), "connections")
I_to_E = Projection(I_net, E_net, connector, I_syn, receptor_type="inhibitory")
print("I --> E\t\t", len(I_to_E), "connections")
input_to_E = Projection(expoisson, E_net, ext_Connector, ext_syn, receptor_type="excitatory")
print("input --> E\t", len(input_to_E), "connections")
print("%d Connecting inhibitory population with connection probability %g, weight %g nA and delay %g ms." % (rank, epsilon, JI, delay))
E_to_I = Projection(E_net, I_net, connector, E_syn, receptor_type="excitatory")
print("E --> I\t\t", len(E_to_I), "connections")
I_to_I = Projection(I_net, I_net, connector, I_syn, receptor_type="inhibitory")
print("I --> I\t\t", len(I_to_I), "connections")
input_to_I = Projection(inpoisson, I_net, ext_Connector, ext_syn, receptor_type="excitatory")
print("input --> I\t", len(input_to_I), "connections")
# read out time used for building
buildCPUTime = timer.elapsedTime()
# === Run simulation ===========================================================
# run, measure computer time
timer.start() # start timer on construction
print("%d Running simulation for %g ms (dt=%sms)." % (rank, simtime, dt))
run(simtime)
print("Done")
simCPUTime = timer.elapsedTime()
# write data to file
#print("%d Writing data to file." % rank)
#(E_net + I_net).write_data("Results/brunel_np%d_%s.pkl" % (np, simulator_name))
if save and not simulator_name=='neuroml':
for pop in [E_net , I_net]:
io = PyNNTextIO(filename="brunel-PyNN-%s-%s-%i.gdf"%(simulator_name, pop.label, rank))
spikes = pop.get_data('spikes', gather=False)
for segment in spikes.segments:
io.write_segment(segment)
io = PyNNTextIO(filename="brunel-PyNN-%s-%s-%i.dat"%(simulator_name, pop.label, rank))
vs = pop.get_data('v', gather=False)
for segment in vs.segments:
io.write_segment(segment)
spike_data = {}
spike_data['senders'] = []
spike_data['times'] = []
index_offset = 1
for pop in [E_net , I_net]:
if rank == 0:
spikes = pop.get_data('spikes', gather=False)
#print(spikes.segments[0].all_data)
num_rec = len(spikes.segments[0].spiketrains)
print("Extracting spike info (%i) for %i cells in %s"%(num_rec,pop.size,pop.label))
#assert(num_rec==len(spikes.segments[0].spiketrains))
for i in range(num_rec):
ss = spikes.segments[0].spiketrains[i]
for s in ss:
index = i+index_offset
#print("Adding spike at %s in %s[%i] (cell %i)"%(s,pop.label,i,index))
spike_data['senders'].append(index)
spike_data['times'].append(s)
index_offset+=pop.size
#from IPython.core.debugger import Tracer
#Tracer()()
E_rate = E_net.mean_spike_count()*1000.0/simtime
I_rate = I_net.mean_spike_count()*1000.0/simtime
# write a short report
nprint("\n--- Brunel Network Simulation ---")
nprint("Nodes : %d" % np)
nprint("Number of Neurons : %d" % N)
nprint("Number of Synapses : %d" % Nsyn)
nprint("Input firing rate : %g" % p_rate)
nprint("Excitatory weight : %g" % JE)
nprint("Inhibitory weight : %g" % JI)
nprint("Excitatory rate : %g Hz" % E_rate)
nprint("Inhibitory rate : %g Hz" % I_rate)
nprint("Build time : %g s" % buildCPUTime)
nprint("Simulation time : %g s" % simCPUTime)
# === Clean up and quit ========================================================
end()
if simulator_name=='neuroml' and jnml_simulator:
from pyneuroml import pynml
lems_file = 'LEMS_Sim_PyNN_NeuroML2_Export.xml'
print('Going to run generated LEMS file: %s on simulator: %s'%(lems_file,jnml_simulator))
if jnml_simulator=='jNeuroML':
results, events = pynml.run_lems_with_jneuroml(lems_file, nogui=True, load_saved_data=True, reload_events=True)
elif jnml_simulator=='jNeuroML_NEURON':
results, events = pynml.run_lems_with_jneuroml_neuron(lems_file, nogui=True, load_saved_data=True, reload_events=True)
spike_data['senders'] = []
spike_data['times'] = []
for k in events.keys():
values = k.split('/')
index = int(values[1]) if values[0]=='E_net' else NE+int(values[1])
n = len(events[k])
print("Loading spikes for %s (index %i): [%s, ..., %s (n=%s)] sec"%(k,index,events[k][0] if n>0 else '-',events[k][-1] if n>0 else '-',n))
for t in events[k]:
spike_data['senders'].append(index)
spike_data['times'].append(t*1000)
#print spike_data
return spike_data
for pop in [
pop_IF_curr_alpha, pop_IF_curr_exp, pop_IF_cond_exp,
pop_IF_cond_alpha, pop_EIF_cond_exp_isfa_ista, pop_HH_cond_exp,
pop_post1, pop_post2
]:
data = pop.get_data('v', gather=False)
filename = "%s_v.dat" % (pop.label)
print("Writing data for %s" % pop)
for segment in data.segments:
vm = segment.analogsignalarrays[0].transpose()[0]
tt = np.array([t * time_step / 1000. for t in range(len(vm))])
times_vm = np.array([tt, vm / 1000.]).transpose()
np.savetxt(filename, times_vm, delimiter='\t', fmt='%s')
filename = "%s.spikes" % (pop.label)
io = PyNNTextIO(filename=filename)
spikes = pop.get_data('spikes', gather=False)
if len(spikes.segments[0].spiketrains) > 0:
print("Writing data for %s" % pop)
for segment in spikes.segments:
io.write_segment(segment)
sim.end()
if '-gui' in sys.argv:
if simulator_name in ['neuron', 'nest', 'brian']:
import matplotlib.pyplot as plt
print("Plotting results of simulation in %s" % simulator_name)
plt.figure("Voltages for IaF cells")
def runBrunelNetwork(g=5.,
eta=2.,
dt=0.1,
simtime=1000.0,
delay=1.5,
epsilon=0.1,
order=2500,
N_rec=50,
N_rec_v=2,
save=False,
simulator_name='nest',
jnml_simulator=None,
extra={}):
exec("from pyNN.%s import *" % simulator_name) in globals()
timer = Timer()
# === Define parameters ========================================================
downscale = 1 # scale number of neurons down by this factor
# scale synaptic weights up by this factor to
# obtain similar dynamics independent of size
order = order # determines size of network:
# 4*order excitatory neurons
# 1*order inhibitory neurons
Nrec = N_rec # number of neurons to record from, per population
epsilon = epsilon # connectivity: proportion of neurons each neuron projects to
# Parameters determining model dynamics, cf Brunel (2000), Figs 7, 8 and Table 1
# here: Case C, asynchronous irregular firing, ~35 Hz
eta = eta # rel rate of external input
g = g # rel strength of inhibitory synapses
J = 0.1 # synaptic weight [mV]
delay = delay # synaptic delay, all connections [ms]
# single neuron parameters
tauMem = 20.0 # neuron membrane time constant [ms]
tauSyn = 0.1 # synaptic time constant [ms]
tauRef = 2.0 # refractory time [ms]
U0 = 0.0 # resting potential [mV]
theta = 20.0 # threshold
# simulation-related parameters
simtime = simtime # simulation time [ms]
dt = dt # simulation step length [ms]
# seed for random generator used when building connections
connectseed = 12345789
use_RandomArray = True # use Python rng rather than NEST rng
# seed for random generator(s) used during simulation
kernelseed = 43210987
# === Calculate derived parameters =============================================
# scaling: compute effective order and synaptic strength
order_eff = int(float(order) / downscale)
J_eff = J * downscale
# compute neuron numbers
NE = int(4 * order_eff) # number of excitatory neurons
NI = int(1 * order_eff) # number of inhibitory neurons
N = NI + NE # total number of neurons
# compute synapse numbers
CE = int(epsilon * NE) # number of excitatory synapses on neuron
CI = int(epsilon * NI) # number of inhibitory synapses on neuron
C = CE + CI # total number of internal synapses per n.
Cext = CE # number of external synapses on neuron
# synaptic weights, scaled for alpha functions, such that
# for constant membrane potential, charge J would be deposited
fudge = 0.00041363506632638 # ensures dV = J at V=0
# excitatory weight: JE = J_eff / tauSyn * fudge
JE = (J_eff / tauSyn) * fudge
# inhibitory weight: JI = - g * JE
JI = -g * JE
# threshold, external, and Poisson generator rates:
nu_thresh = theta / (J_eff * CE * tauMem)
nu_ext = eta * nu_thresh # external rate per synapse
p_rate = 1000 * nu_ext * Cext # external input rate per neuron (Hz)
# number of synapses---just so we know
Nsyn = (
C + 1
) * N + 2 * Nrec # number of neurons * (internal synapses + 1 synapse from PoissonGenerator) + 2synapses" to spike detectors
# put cell parameters into a dict
cell_params = {
'tau_m': tauMem,
'tau_syn_E': tauSyn,
'tau_syn_I': tauSyn,
'tau_refrac': tauRef,
'v_rest': U0,
'v_reset': U0,
'v_thresh': theta,
'cm': 0.001
} # (nF)
# === Build the network ========================================================
# clear all existing network elements and set resolution and limits on delays.
# For NEST, limits must be set BEFORE connecting any elements
#extra = {'threads' : 2}
rank = setup(timestep=dt, max_delay=delay, **extra)
print("rank =", rank)
np = num_processes()
print("np =", np)
import socket
host_name = socket.gethostname()
print("Host #%d is on %s" % (rank + 1, host_name))
if 'threads' in extra:
print("%d Initialising the simulator with %d threads..." %
(rank, extra['threads']))
else:
print("%d Initialising the simulator with single thread..." % rank)
# Small function to display information only on node 1
def nprint(s):
if rank == 0:
print(s)
timer.start() # start timer on construction
print("%d Setting up random number generator" % rank)
rng = NumpyRNG(kernelseed, parallel_safe=True)
print("%d Creating excitatory population with %d neurons." % (rank, NE))
celltype = IF_curr_alpha(**cell_params)
celltype.default_initial_values[
'v'] = U0 # Setting default init v, useful for NML2 export
E_net = Population(NE, celltype, label="E_net")
print("%d Creating inhibitory population with %d neurons." % (rank, NI))
I_net = Population(NI, celltype, label="I_net")
print(
"%d Initialising membrane potential to random values between %g mV and %g mV."
% (rank, U0, theta))
uniformDistr = RandomDistribution('uniform', low=U0, high=theta, rng=rng)
E_net.initialize(v=uniformDistr)
I_net.initialize(v=uniformDistr)
print("%d Creating excitatory Poisson generator with rate %g spikes/s." %
(rank, p_rate))
source_type = SpikeSourcePoisson(rate=p_rate)
expoisson = Population(NE, source_type, label="expoisson")
print("%d Creating inhibitory Poisson generator with the same rate." %
rank)
inpoisson = Population(NI, source_type, label="inpoisson")
# Record spikes
print("%d Setting up recording in excitatory population." % rank)
E_net.record('spikes')
if N_rec_v > 0:
E_net[0:min(NE, N_rec_v)].record('v')
print("%d Setting up recording in inhibitory population." % rank)
I_net.record('spikes')
if N_rec_v > 0:
I_net[0:min(NI, N_rec_v)].record('v')
progress_bar = ProgressBar(width=20)
connector = FixedProbabilityConnector(epsilon,
rng=rng,
callback=progress_bar)
E_syn = StaticSynapse(weight=JE, delay=delay)
I_syn = StaticSynapse(weight=JI, delay=delay)
ext_Connector = OneToOneConnector(callback=progress_bar)
ext_syn = StaticSynapse(weight=JE, delay=dt)
print(
"%d Connecting excitatory population with connection probability %g, weight %g nA and delay %g ms."
% (rank, epsilon, JE, delay))
E_to_E = Projection(E_net,
E_net,
connector,
E_syn,
receptor_type="excitatory")
print("E --> E\t\t", len(E_to_E), "connections")
I_to_E = Projection(I_net,
E_net,
connector,
I_syn,
receptor_type="inhibitory")
print("I --> E\t\t", len(I_to_E), "connections")
input_to_E = Projection(expoisson,
E_net,
ext_Connector,
ext_syn,
receptor_type="excitatory")
print("input --> E\t", len(input_to_E), "connections")
print(
"%d Connecting inhibitory population with connection probability %g, weight %g nA and delay %g ms."
% (rank, epsilon, JI, delay))
E_to_I = Projection(E_net,
I_net,
connector,
E_syn,
receptor_type="excitatory")
print("E --> I\t\t", len(E_to_I), "connections")
I_to_I = Projection(I_net,
I_net,
connector,
I_syn,
receptor_type="inhibitory")
print("I --> I\t\t", len(I_to_I), "connections")
input_to_I = Projection(inpoisson,
I_net,
ext_Connector,
ext_syn,
receptor_type="excitatory")
print("input --> I\t", len(input_to_I), "connections")
# read out time used for building
buildCPUTime = timer.elapsedTime()
# === Run simulation ===========================================================
# run, measure computer time
timer.start() # start timer on construction
print("%d Running simulation for %g ms (dt=%sms)." % (rank, simtime, dt))
run(simtime)
print("Done")
simCPUTime = timer.elapsedTime()
# write data to file
#print("%d Writing data to file." % rank)
#(E_net + I_net).write_data("Results/brunel_np%d_%s.pkl" % (np, simulator_name))
if save and not simulator_name == 'neuroml':
for pop in [E_net, I_net]:
io = PyNNTextIO(filename="brunel-PyNN-%s-%s-%i.gdf" %
(simulator_name, pop.label, rank))
spikes = pop.get_data('spikes', gather=False)
for segment in spikes.segments:
io.write_segment(segment)
io = PyNNTextIO(filename="brunel-PyNN-%s-%s-%i.dat" %
(simulator_name, pop.label, rank))
vs = pop.get_data('v', gather=False)
for segment in vs.segments:
io.write_segment(segment)
spike_data = {}
spike_data['senders'] = []
spike_data['times'] = []
index_offset = 1
for pop in [E_net, I_net]:
if rank == 0:
spikes = pop.get_data('spikes', gather=False)
#print(spikes.segments[0].all_data)
num_rec = len(spikes.segments[0].spiketrains)
print("Extracting spike info (%i) for %i cells in %s" %
(num_rec, pop.size, pop.label))
#assert(num_rec==len(spikes.segments[0].spiketrains))
for i in range(num_rec):
ss = spikes.segments[0].spiketrains[i]
for s in ss:
index = i + index_offset
#print("Adding spike at %s in %s[%i] (cell %i)"%(s,pop.label,i,index))
spike_data['senders'].append(index)
spike_data['times'].append(s)
index_offset += pop.size
#from IPython.core.debugger import Tracer
#Tracer()()
E_rate = E_net.mean_spike_count() * 1000.0 / simtime
I_rate = I_net.mean_spike_count() * 1000.0 / simtime
# write a short report
nprint("\n--- Brunel Network Simulation ---")
nprint("Nodes : %d" % np)
nprint("Number of Neurons : %d" % N)
nprint("Number of Synapses : %d" % Nsyn)
nprint("Input firing rate : %g" % p_rate)
nprint("Excitatory weight : %g" % JE)
nprint("Inhibitory weight : %g" % JI)
nprint("Excitatory rate : %g Hz" % E_rate)
nprint("Inhibitory rate : %g Hz" % I_rate)
nprint("Build time : %g s" % buildCPUTime)
nprint("Simulation time : %g s" % simCPUTime)
# === Clean up and quit ========================================================
end()
if simulator_name == 'neuroml' and jnml_simulator:
from pyneuroml import pynml
lems_file = 'LEMS_Sim_PyNN_NeuroML2_Export.xml'
print('Going to run generated LEMS file: %s on simulator: %s' %
(lems_file, jnml_simulator))
if jnml_simulator == 'jNeuroML':
results, events = pynml.run_lems_with_jneuroml(
lems_file,
nogui=True,
load_saved_data=True,
reload_events=True)
elif jnml_simulator == 'jNeuroML_NEURON':
results, events = pynml.run_lems_with_jneuroml_neuron(
lems_file,
nogui=True,
load_saved_data=True,
reload_events=True)
spike_data['senders'] = []
spike_data['times'] = []
for k in events.keys():
values = k.split('/')
index = int(
values[1]) if values[0] == 'E_net' else NE + int(values[1])
n = len(events[k])
print(
"Loading spikes for %s (index %i): [%s, ..., %s (n=%s)] sec" %
(k, index, events[k][0] if n > 0 else '-',
events[k][-1] if n > 0 else '-', n))
for t in events[k]:
spike_data['senders'].append(index)
spike_data['times'].append(t * 1000)
#print spike_data
return spike_data
def runBrunelNetwork(
g=5.0,
eta=2.0,
dt=0.1,
simtime=1000.0,
delay=1.5,
epsilon=0.1,
order=2500,
N_rec=50,
N_rec_v=2,
save=False,
simulator_name="nest",
extra={},
):
exec("from pyNN.%s import *" % simulator_name) in globals()
timer = Timer()
# === Define parameters ========================================================
downscale = 1 # scale number of neurons down by this factor
# scale synaptic weights up by this factor to
# obtain similar dynamics independent of size
order = order # determines size of network:
# 4*order excitatory neurons
# 1*order inhibitory neurons
Nrec = N_rec # number of neurons to record from, per population
epsilon = epsilon # connectivity: proportion of neurons each neuron projects to
# Parameters determining model dynamics, cf Brunel (2000), Figs 7, 8 and Table 1
# here: Case C, asynchronous irregular firing, ~35 Hz
eta = eta # rel rate of external input
g = g # rel strength of inhibitory synapses
J = 0.1 # synaptic weight [mV]
delay = delay # synaptic delay, all connections [ms]
# single neuron parameters
tauMem = 20.0 # neuron membrane time constant [ms]
tauSyn = 0.1 # synaptic time constant [ms]
tauRef = 2.0 # refractory time [ms]
U0 = 0.0 # resting potential [mV]
theta = 20.0 # threshold
# simulation-related parameters
simtime = simtime # simulation time [ms]
dt = dt # simulation step length [ms]
# seed for random generator used when building connections
connectseed = 12345789
use_RandomArray = True # use Python rng rather than NEST rng
# seed for random generator(s) used during simulation
kernelseed = 43210987
# === Calculate derived parameters =============================================
# scaling: compute effective order and synaptic strength
order_eff = int(float(order) / downscale)
J_eff = J * downscale
# compute neuron numbers
NE = int(4 * order_eff) # number of excitatory neurons
NI = int(1 * order_eff) # number of inhibitory neurons
N = NI + NE # total number of neurons
# compute synapse numbers
CE = int(epsilon * NE) # number of excitatory synapses on neuron
CI = int(epsilon * NI) # number of inhibitory synapses on neuron
C = CE + CI # total number of internal synapses per n.
Cext = CE # number of external synapses on neuron
# synaptic weights, scaled for alpha functions, such that
# for constant membrane potential, charge J would be deposited
fudge = 0.00041363506632638 # ensures dV = J at V=0
# excitatory weight: JE = J_eff / tauSyn * fudge
JE = (J_eff / tauSyn) * fudge
# inhibitory weight: JI = - g * JE
JI = -g * JE
# threshold, external, and Poisson generator rates:
nu_thresh = theta / (J_eff * CE * tauMem)
nu_ext = eta * nu_thresh # external rate per synapse
p_rate = 1000 * nu_ext * Cext # external input rate per neuron (Hz)
# number of synapses---just so we know
Nsyn = (
C + 1
) * N + 2 * Nrec # number of neurons * (internal synapses + 1 synapse from PoissonGenerator) + 2synapses" to spike detectors
# put cell parameters into a dict
cell_params = {
"tau_m": tauMem,
"tau_syn_E": tauSyn,
"tau_syn_I": tauSyn,
"tau_refrac": tauRef,
"v_rest": U0,
"v_reset": U0,
"v_thresh": theta,
"cm": 0.001,
} # (nF)
# === Build the network ========================================================
# clear all existing network elements and set resolution and limits on delays.
# For NEST, limits must be set BEFORE connecting any elements
# extra = {'threads' : 2}
rank = setup(timestep=dt, max_delay=delay, **extra)
print("rank =", rank)
np = num_processes()
print("np =", np)
import socket
host_name = socket.gethostname()
print("Host #%d is on %s" % (rank + 1, host_name))
if "threads" in extra:
print("%d Initialising the simulator with %d threads..." % (rank, extra["threads"]))
else:
print("%d Initialising the simulator with single thread..." % rank)
# Small function to display information only on node 1
def nprint(s):
if rank == 0:
print(s)
timer.start() # start timer on construction
print("%d Setting up random number generator" % rank)
rng = NumpyRNG(kernelseed, parallel_safe=True)
print("%d Creating excitatory population with %d neurons." % (rank, NE))
celltype = IF_curr_alpha(**cell_params)
E_net = Population(NE, celltype, label="E_net")
print("%d Creating inhibitory population with %d neurons." % (rank, NI))
I_net = Population(NI, celltype, label="I_net")
print("%d Initialising membrane potential to random values between %g mV and %g mV." % (rank, U0, theta))
uniformDistr = RandomDistribution("uniform", low=U0, high=theta, rng=rng)
E_net.initialize(v=uniformDistr)
I_net.initialize(v=uniformDistr)
print("%d Creating excitatory Poisson generator with rate %g spikes/s." % (rank, p_rate))
source_type = SpikeSourcePoisson(rate=p_rate)
expoisson = Population(NE, source_type, label="expoisson")
print("%d Creating inhibitory Poisson generator with the same rate." % rank)
inpoisson = Population(NI, source_type, label="inpoisson")
# Record spikes
print("%d Setting up recording in excitatory population." % rank)
E_net.sample(Nrec).record("spikes")
E_net[0 : min(NE, N_rec_v)].record("v")
print("%d Setting up recording in inhibitory population." % rank)
I_net.sample(Nrec).record("spikes")
I_net[0 : min(NI, N_rec_v)].record("v")
progress_bar = ProgressBar(width=20)
connector = FixedProbabilityConnector(epsilon, rng=rng, callback=progress_bar)
E_syn = StaticSynapse(weight=JE, delay=delay)
I_syn = StaticSynapse(weight=JI, delay=delay)
ext_Connector = OneToOneConnector(callback=progress_bar)
ext_syn = StaticSynapse(weight=JE, delay=dt)
print(
"%d Connecting excitatory population with connection probability %g, weight %g nA and delay %g ms."
% (rank, epsilon, JE, delay)
)
E_to_E = Projection(E_net, E_net, connector, E_syn, receptor_type="excitatory")
print("E --> E\t\t", len(E_to_E), "connections")
I_to_E = Projection(I_net, E_net, connector, I_syn, receptor_type="inhibitory")
print("I --> E\t\t", len(I_to_E), "connections")
input_to_E = Projection(expoisson, E_net, ext_Connector, ext_syn, receptor_type="excitatory")
print("input --> E\t", len(input_to_E), "connections")
print(
"%d Connecting inhibitory population with connection probability %g, weight %g nA and delay %g ms."
% (rank, epsilon, JI, delay)
)
E_to_I = Projection(E_net, I_net, connector, E_syn, receptor_type="excitatory")
print("E --> I\t\t", len(E_to_I), "connections")
I_to_I = Projection(I_net, I_net, connector, I_syn, receptor_type="inhibitory")
print("I --> I\t\t", len(I_to_I), "connections")
input_to_I = Projection(inpoisson, I_net, ext_Connector, ext_syn, receptor_type="excitatory")
print("input --> I\t", len(input_to_I), "connections")
# read out time used for building
buildCPUTime = timer.elapsedTime()
# === Run simulation ===========================================================
# run, measure computer time
timer.start() # start timer on construction
print("%d Running simulation for %g ms." % (rank, simtime))
run(simtime)
simCPUTime = timer.elapsedTime()
# write data to file
# print("%d Writing data to file." % rank)
# (E_net + I_net).write_data("Results/brunel_np%d_%s.pkl" % (np, simulator_name))
if save:
for pop in [E_net, I_net]:
io = PyNNTextIO(filename="brunel-PyNN-%s-%s-%i.gdf" % (simulator_name, pop.label, rank))
spikes = pop.get_data("spikes", gather=False)
for segment in spikes.segments:
io.write_segment(segment)
io = PyNNTextIO(filename="brunel-PyNN-%s-%s-%i.dat" % (simulator_name, pop.label, rank))
vs = pop.get_data("v", gather=False)
for segment in vs.segments:
io.write_segment(segment)
spike_data = {}
spike_data["senders"] = []
spike_data["times"] = []
index_offset = 0
for pop in [E_net, I_net]:
spikes = pop.get_data("spikes", gather=False)
# print(spikes.segments[0].all_data)
num_rec = min(pop.size, N_rec)
print("Extracting spike info (%i) for %i cells in %s" % (num_rec, pop.size, pop.label))
assert num_rec == len(spikes.segments[0].spiketrains)
for i in range(num_rec):
ss = spikes.segments[0].spiketrains[i]
for s in ss:
index = i + index_offset
# print("Adding spike at %s in %s[%i] (cell %i)"%(s,pop.label,i,index))
spike_data["senders"].append(index)
spike_data["times"].append(s)
index_offset += pop.size
# from IPython.core.debugger import Tracer
# Tracer()()
E_rate = E_net.mean_spike_count() * 1000.0 / simtime
I_rate = I_net.mean_spike_count() * 1000.0 / simtime
# write a short report
nprint("\n--- Brunel Network Simulation ---")
nprint("Nodes : %d" % np)
nprint("Number of Neurons : %d" % N)
nprint("Number of Synapses : %d" % Nsyn)
nprint("Input firing rate : %g" % p_rate)
nprint("Excitatory weight : %g" % JE)
nprint("Inhibitory weight : %g" % JI)
nprint("Excitatory rate : %g Hz" % E_rate)
nprint("Inhibitory rate : %g Hz" % I_rate)
nprint("Build time : %g s" % buildCPUTime)
nprint("Simulation time : %g s" % simCPUTime)
# === Clean up and quit ========================================================
end()
return spike_data
