def pif_reset():
defaultclock.reinit()
sim = Network()
I = 0.2*nA
R = 1*Mohm
lifeq = """
dV/dt = I*R/ms : volt
Vth : volt
"""
thstep = 15*mV
nrn = NeuronGroup(1, lifeq, threshold="V>=Vth", reset="V=0*mV")
nrn.V = 0*mV
nrn.Vth = thstep
sim.add(nrn)
#connection = Connection(inputgrp, nrn, state="V", weight=0.5*mV)
#sim.add(inputgrp, connection)
vmon = StateMonitor(nrn, "V", record=True)
thmon = StateMonitor(nrn, "Vth", record=True)
spikemon = SpikeMonitor(nrn, record=True)
sim.add(vmon, thmon, spikemon)
sim.run(duration)
return vmon, thmon, spikemon
def runsim(Nin, weight, fout, sync):
sim = Network()
clear(True)
gc.collect()
defaultclock.reinit()
duration = 5*second
lifeq = "dV/dt = -V/(10*ms) : volt"
nrndef = {"model": lifeq, "threshold": "V>=15*mV", "reset": "V=0*mV",
"refractory": 2*ms}
fin = load_or_calibrate(nrndef, Nin, weight, sync, fout,
Vth=15*mV, tau=10*ms)
# print("Calibrated frequencies:")
# print(", ".join(str(f) for f in fin))
inputgroups = []
connections = []
neurons = []
Nneurons = len(fin)
neurons = NeuronGroup(Nneurons, **nrndef)
for idx in range(Nneurons):
fin_i = fin[idx]
sync_i, sigma_i = sync[idx]
inputgrp = sl.tools.fast_synchronous_input_gen(Nin, fin_i,
sync_i, sigma_i,
duration)
defaultclock.reinit()
conn = Connection(inputgrp, neurons[idx], state="V", weight=weight)
inputgroups.append(inputgrp)
connections.append(conn)
voltagemon = StateMonitor(neurons, "V", record=True)
spikemon = SpikeMonitor(neurons, record=True)
sim.add(neurons, voltagemon, spikemon)
sim.add(*inputgroups)
sim.add(*connections)
print("Running {} {} {}".format(Nin, weight, fout))
sim.run(duration, report="stdout")
mnpss = []
allnpss = []
for idx in range(Nneurons):
vmon = voltagemon[idx]
smon = spikemon[idx]
# print("Desired firing rate: {}".format(fout))
# print("Actual firing rate: {}".format(len(smon)/duration))
if len(smon) > 0:
npss = sl.tools.npss(vmon, smon, 0*mV, 15*mV, 10*ms, 2*ms)
else:
npss = 0
mnpss.append(np.mean(npss))
allnpss.append(npss)
nrndeftuple = tuple(nrndef.items())
key = (nrndeftuple, Nin, weight, tuple(sync), fout, 15*mV, 10*ms)
save_data(key, allnpss)
imshape = (len(sigma), len(Sin))
imextent = (0, 1, 0, 4.0)
mnpss = np.reshape(mnpss, imshape, order="F")
plt.figure()
plt.imshow(mnpss, aspect="auto", origin="lower", extent=imextent,
interpolation="none", vmin=0, vmax=1)
cbar = plt.colorbar()
cbar.set_label("$\overline{M}$")
plt.xlabel("$S_{in}$")
plt.ylabel("$\sigma_{in}$ (ms)")
filename = "npss_{}_{}_{}".format(Nin, weight, fout).replace(".", "")
plt.savefig(filename+".pdf")
plt.savefig(filename+".png")
print("{} saved".format(filename))
voltages = voltagemon.values
spiketrains = spikemon.spiketimes.values()
pickle.dump({"voltages": voltages, "spiketrains": spiketrains},
open(filename+".pkl", 'w'))
return voltagemon, spikemon
def runsim(fin):
clear(True)
gc.collect()
defaultclock.reinit()
weight = 0.16*mV
sim = Network()
duration = 2.0*second
Vth = 15*mV
Vreset = 13.65*mV
trefr = 2*ms
lifeq = """
dV/dt = -V/(10*ms) : volt
Vth : volt
"""
nrndef = {"model": lifeq, "threshold": "V>=Vth", "reset": "V=Vreset",
"refractory": 0.1*ms}
inputgroups = []
connections = []
neurons = []
Nneurons = len(fin)
neurons = NeuronGroup(Nneurons, **nrndef)
neurons.V = 0*mV
neurons.Vth = 15*mV
for idx in range(Nneurons):
fin_i = fin[idx]*Hz
inputgrp = PoissonGroup(50, fin_i)
conn = Connection(inputgrp, neurons[idx], state="V", weight=weight)
inputgroups.append(inputgrp)
connections.append(conn)
voltagemon = StateMonitor(neurons, "V", record=True)
spikemon = SpikeMonitor(neurons, record=True)
sim.add(neurons, voltagemon, spikemon)
sim.add(*inputgroups)
sim.add(*connections)
@network_operation
def refractory_threshold(clock):
for idx in range(Nneurons):
if (len(spikemon.spiketimes[idx])
and clock.t < spikemon.spiketimes[idx][-1]*second+trefr):
neurons.Vth[idx] = 100*mV
else:
neurons.Vth[idx] = Vth
sim.add(refractory_threshold)
print("Running simulation of {} neurons for {} s".format(Nneurons, duration))
sim.run(duration, report="stdout")
mnpss = []
allnpss = []
outisi = []
for idx in range(Nneurons):
vmon = voltagemon[idx]
smon = spikemon[idx]
if not len(smon):
continue
outisi.append(duration*1000/len(smon))
if len(smon) > 0:
npss = sl.tools.npss(vmon, smon, 0*mV, 15*mV, 10*ms, 2*ms)
else:
npss = 0
mnpss.append(np.mean(npss))
allnpss.append(npss)
return outisi, mnpss
