def exercise4_curve(tau_refrac):
hugs = p.Population(1, cellclass=p.IF_cond_exp)
hugs.set("tau_refrac", tau_refrac)
start = 100.
stop = 1100.
frequency = []
currentSpace = linspace(0.1,5.0,50)
for current in currentSpace:
hugs.record()
hugs.record_v()
dcsource = p.DCSource(amplitude=current, start=start, stop=stop)
dcsource.inject_into(hugs)
p.run(1100.)
spikes = hugs.getSpikes()
frequency.append(len(spikes) / (stop - start))
print(current, frequency[-1])
p.reset()
return currentSpace, frequency
def sim(trial):
global direction
global inspikes
global outspikes
global outspikes2
global outvolts
direction=0
#print direction
p.reset()
#for direction in dirs:
for n in range(len(neuron)):
neuron[n].set('spike_times',DATA[direction][n][trial])
p.run(2000)
outspikes=[0,0]
outspikes2=[]
outvolts=[]
for i,o in enumerate(out):
outspikes[i]=o.get_spike_counts().values()[0]
outspikes2.append(o.getSpikes())
outvolts.append(o.get_v())
inspikes=[0,0,0,0,0]
for i,n in enumerate(neuron):
inspikes[i]=n.get_spike_counts().values()[0]
"""
fig = figure()
ax = fig.add_subplot(1,1,1)
hold(True)
for i in range(nout):
ax.plot(outspikes2[i][:,1],i*ones_like(outspikes2[i][:,1]),'b|')
ax.set_ylim(-6,5)
for i in range(nneurons):
# ax.plot(DATA[direction][i][trial],-1-i*ones_like(DATA[direction][i][trial]),'r|')
inspikes2=neuron[i].getSpikes()
ax.plot(inspikes2,-1-i*ones_like(inspikes2),'r|')
#ax2=fig.add_subplot(2,1,2)
#ax2.plot(outvolts[0][:,1],outvolts[0][:,2])
"""
return inspikes,outspikes
def exercise2():
hugs = p.Population(1, cellclass=p.HH_cond_exp)
start = 100.0
stop = 1100.0
frequency = []
currentSpace = linspace(0.1, 10, 100)
for current in currentSpace:
hugs.record()
hugs.record_v()
dcsource = p.DCSource(amplitude=current, start=start, stop=stop)
dcsource.inject_into(hugs)
p.run(1100.0)
spikes = hugs.getSpikes()
frequency.append(len(spikes) / (stop - start))
p.reset()
plot(currentSpace, frequency)
xlabel("Current/nA")
ylabel("Frequency/kHz")
savefig("../output/day4_figure2.png")
show()
def subplot(v_rest = -65.0, out = '../output/day5figure1.png'):
figure(figsize=(12, 10))
suptitle('Threshold: ' + str(v_rest) +'mV')
epspPot, econdGsyn, epspSpikes = exercise1(v_rest)
p.reset()
p.reset()
ipspPot, icondGsyn, ipspSpikes = exercise2(v_rest)
plt.subplot(2,2,1)
plot(epspPot[:,1], epspPot[:,2])
plot(epspSpikes[:,1], [max(epspPot[:,2])]*len(epspSpikes), '|', color='r', markersize=20.0)
xlabel('Time/ms')
ylabel('Membrane potential/mV')
title('EPSPs after excitatory postsynaptic signal')
plt.subplot(2,2,2)
plot(ipspPot[:,1], ipspPot[:,2])
plot(ipspSpikes[:,1], [max(ipspPot[:,2])]*len(ipspSpikes), '|', color='r', markersize=20.0)
xlabel('Time/ms')
ylabel('Membrane potential/mV')
title('IPSPs after inhibitory presynaptic signal')
plt.subplot(2,2,3)
plot(econdGsyn[:,1], econdGsyn[:,2])
plot(epspSpikes[:,1], [max(econdGsyn[:,2])]*len(epspSpikes), '|', color='r', markersize=20.0)
xlabel('Time/ms')
ylabel(u'Excitatory conductance/\u00B5S')
plt.subplot(2,2,4)
plot(icondGsyn[:,1], icondGsyn[:,3])
plot(ipspSpikes[:,1], [max(icondGsyn[:,3])]*len(ipspSpikes), '|', color='r',markersize=20.0)
xlabel('Time/ms')
ylabel(u'Inhibitory conductance/\u00B5S')
savefig(out)
draw()
show()
def exercise2():
figure(figsize=(10,6))
hugs = p.Population(1, cellclass=p.HH_cond_exp)
start = 100.
stop = 1100.
frequency = []
currentSpace = linspace(0.1,10,100)
for current in currentSpace:
hugs.record()
hugs.record_v()
dcsource = p.DCSource(amplitude=current, start=start, stop=stop)
dcsource.inject_into(hugs)
p.run(1100.)
spikes = hugs.getSpikes()
frequency.append(len(spikes) / (stop - start))
p.reset()
plot(currentSpace, frequency)
title('Current-frequency diagram for a HH model neuron')
xlabel('Current/nA')
ylabel('Frequency/kHz')
savefig('../output/day4_figure2.png')
show()
def exercise5(tau_m, v_thresh=-55.0):
hugs = p.Population(1, cellclass=p.IF_cond_exp)
hugs.set("tau_refrac", 2.0)
hugs.set("tau_m", tau_m)
hugs.set("v_thresh", v_thresh)
start = 100.0
stop = 400.0
dcsource = p.DCSource(amplitude=0.95, start=start, stop=stop)
dcsource.inject_into(hugs)
hugs.record()
hugs.record_v()
p.run(500.0)
pot = hugs.get_v()
spikes = hugs.getSpikes()
p.reset()
return pot, spikes, (stop - start)
import pyNN.neuron as sim # can of course replace `nest` with `neuron`, `brian`, etc.
import matplotlib.pyplot as plt
from quantities import nA
sim.setup()
cell = sim.Population(1, sim.HH_cond_exp())
step_current = sim.DCSource(start=20.0, stop=80.0)
step_current.inject_into(cell)
cell.record('v')
for amp in (-0.2, -0.1, 0.0, 0.1, 0.2):
step_current.amplitude = amp
sim.run(100.0)
sim.reset(annotations={"amplitude": amp * nA})
data = cell.get_data()
sim.end()
for segment in data.segments:
vm = segment.analogsignals[0]
plt.plot(vm.times, vm,
label=str(segment.annotations["amplitude"]))
plt.legend(loc="upper left")
plt.xlabel("Time (%s)" % vm.times.units._dimensionality)
plt.ylabel("Membrane potential (%s)" % vm.units._dimensionality)
plt.show()
import scipy.io as io
from pylab import *
import numpy as np
import pyNN.neuron as p
p.setup(timestep=0.1)
p.reset()
import NeuroTools.stgen as nts
# initialise DATA
data1 = io.loadmat('/home/bude/k0018000/NEUROprojekt/bootstrap_joe093-3-C3-MO.mat')['GDFcell'][0]
data2 = io.loadmat('/home/bude/k0018000/NEUROprojekt/bootstrap_joe108-7-C3-MO.mat')['GDFcell'][0]
data3 = io.loadmat('/home/bude/k0018000/NEUROprojekt/bootstrap_joe112-5-C3-MO.mat')['GDFcell'][0]
data4 = io.loadmat('/home/bude/k0018000/NEUROprojekt/bootstrap_joe112-6-C3-MO.mat')['GDFcell'][0]
data5 = io.loadmat('/home/bude/k0018000/NEUROprojekt/bootstrap_joe145-4-C3-MO.mat')['GDFcell'][0]
data = [data1, data2, data3, data4, data5]
ndirections = 6
nneurons = 5
ntrials = 30
DATA = [[[array([]) for i in range(ntrials)] for i in range(nneurons)] for i in range(ndirections)]
for i in range(ndirections):
for j in range(nneurons):
for k in range(ntrials):
DATA[i][j][k]=data[j][i][data[j][i][:,0]==k+1,1]
DATA[i][j][k].sort()
import pyNN.neuron as sim # can of course replace `nest` with `neuron`, `brian`, etc.
import matplotlib.pyplot as plt
from quantities import nA
sim.setup()
cell = sim.Population(1, sim.HH_cond_exp())
step_current = sim.DCSource(start=20.0, stop=80.0)
step_current.inject_into(cell)
cell.record('v')
for amp in (-0.2, -0.1, 0.0, 0.1, 0.2):
step_current.amplitude = amp
sim.run(100.0)
sim.reset(annotations={"amplitude": amp * nA})
data = cell.get_data()
sim.end()
for segment in data.segments:
vm = segment.analogsignals[0]
plt.plot(vm.times, vm, label=str(segment.annotations["amplitude"]))
plt.legend(loc="upper left")
plt.xlabel("Time (%s)" % vm.times.units._dimensionality)
plt.ylabel("Membrane potential (%s)" % vm.units._dimensionality)
plt.show()
def simulate():
global direction
global inspikes
global outspikes
global outspikes2
global outvolts
global prj
print "Simulating..."
# set inputs
for ntrial in range(ntrials):
for direction in dirs:
print "Training direction:", direction, ", trial:", ntrial+1
print "Reading data:"
trainingset = [DATA[direction][nneuron][ntrial] for nneuron in range(nneurons)]
trainingset = [trainingset[i][startstimulus < trainingset[i]] for i in range(nneurons)]
trainingset = [trainingset[i][trainingset[i] < endstimulus] for i in range(nneurons)]
# run simulation
p.reset()
for o in out:
o.record()
o.record_v()
for i, neuron in enumerate(neurons):
neuron.set('spike_times', trainingset[i])
#neuron[nneuron].set('spike_times',arange(1,1901,100))
neuron.record()
p.run(2000)
outSpikeTimes = [[] for i in range(nout)]
outvolts = [[] for i in range(nout)]
## plot spike trains
#fig = figure()
#hold(True)
#ax = fig.add_subplot(1,1,1)
#title("Direction "+str(direction)+", Trial "+str(ntrial+1))
for j, o in enumerate(out):
spikes = list(o.getSpikes()[:,1])
#print j, spikes, len(spikes), type(spikes)
outSpikeTimes[j] = spikes
outvolts[j] = o.get_v()
print "--------------------------------"
#print j, outSpikeTimes[j], len(outSpikeTimes[j])
#ax.plot(outSpikeTimes[j], [j]*len(outSpikeTimes[j]), 'b|', markersize = 20.)
inspikes=[0 for i in range(nneurons)]
outspikes=[0 for i in range(nout)]
outspikes2=[]
outvolts=[]
for i,o in enumerate(out):
outspikes[i] = o.get_spike_counts().values()[0]
#outspikes2.append(o.getSpikes())
outvolts.append(o.get_v())
for i,neuron in enumerate(neurons):
inspikes[i] = neurons[i].get_spike_counts().values()[0]
#print inspikes[i]
def learn():
global direction
global inspikes
global outspikes
global outspikes2
global outvolts
print "Learning..."
#-----------------------------------------------------
def updateWeights():
global direction
global inspikes
global outspikes
global outspikes2
global outvolts
global prj
print "outspikes:", outspikes
print "Updating..."
adjust=0.02
negadjust=0.02
nmax=inspikes.index(max(inspikes))
if (outspikes[0]<outspikes[1]) and (direction==dirs[1]):
prj[2*nmax+1].setWeights(prj[2*nmax+1].getWeights()[0]+adjust)
print 'correct 0'
print "Updated to:", prj[2*nmax+1].getWeights()
elif (outspikes[0]>outspikes[1]) and (direction==dirs[0]):
prj[2*nmax+0].setWeights(prj[2*nmax+0].getWeights()[0]+adjust)
print 'correct 3'
print "Updated to:", prj[2*nmax+0].getWeights()
elif (outspikes[0]>=outspikes[1]) and (direction==dirs[1]):
print 'wrong 0'
prj[2*nmax+0].setWeights(max(0,prj[2*nmax+0].getWeights()[0]-negadjust))
print "Updated to:", prj[2*nmax+0].getWeights()
elif (outspikes[0]<=outspikes[1]) and (direction==dirs[0]):
prj[2*nmax+1].setWeights(max(0,prj[2*nmax+1].getWeights()[0]-negadjust))
print 'wrong 3'
print "Updated to:", prj[2*nmax+1].getWeights()
else:
print 'no 5'
print
updateWeights()
#-----------------------------------------------------
for p in prj:
currentWeight = p.getWeights()[0]
print direction, ntrial, p, currentWeight
learn()
goodWeights = [pr.getWeights() for i, pr in enumerate(prj)]
print "goodWeights:", goodWeights
pickle.dump(goodWeights, open("goodWeightsFile.pkl", "wb"))
def validate():
global direction
global inspikes
global outspikes
global outspikes2
global outvolts
global prj
prj = []
for i in range(nneurons):
for j in range(nout):
prj.append(p.Projection(neurons[i], out[j], target="excitatory",method=p.AllToAllConnector()))
prj[-1].setWeights(initweight)
goodWeights = pickle.load(open("goodWeightsFile.pkl"))
for i, pr in enumerate(prj):
print goodWeights[i], pr
pr.setWeights(goodWeights[i])
print "Validating..."
# set inputs
for ntrial in range(ntrials, ntrials+nindependents):
for direction in dirs:
print "Checking direction:", direction, ", trial:", ntrial+1
print "Reading data:"
trainingset = [DATA[direction][nneuron][ntrial] for nneuron in range(nneurons)]
trainingset = [trainingset[i][startstimulus < trainingset[i]] for i in range(nneurons)]
trainingset = [trainingset[i][trainingset[i] < endstimulus] for i in range(nneurons)]
#import pdb; pdb.set_trace()
# run simulation
p.reset()
for o in out:
o.record()
o.record_v()
for i, neuron in enumerate(neurons):
neuron.set('spike_times', trainingset[i])
#neuron[nneuron].set('spike_times',arange(1,1901,100))
neuron.record()
p.run(2000)
outSpikeTimes = [[] for i in range(nout)]
outvolts = [[] for i in range(nout)]
## plot spike trains
#fig = figure()
#hold(True)
#ax = fig.add_subplot(1,1,1)
#title("Direction "+str(direction)+", Trial "+str(ntrial+1))
for j, o in enumerate(out):
spikes = list(o.getSpikes()[:,1])
#print j, spikes, len(spikes), type(spikes)
outSpikeTimes[j] = spikes
outvolts[j] = o.get_v()
#print j, outSpikeTimes[j], len(outSpikeTimes[j])
#ax.plot(outSpikeTimes[j], [j]*len(outSpikeTimes[j]), 'b|', markersize = 20.)
inspikes=[0 for i in range(nneurons)]
outspikes=[0 for i in range(nout)]
outspikes2=[]
outvolts=[]
for i,o in enumerate(out):
outspikes[i] = o.get_spike_counts().values()[0]
#outspikes2.append(o.getSpikes())
outvolts.append(o.get_v())
print "outspikes:", outspikes
for i,neuron in enumerate(neurons):
inspikes[i] = neurons[i].get_spike_counts().values()[0]
#print inspikes[i]
def check():
global direction
global inspikes
global outspikes
global outspikes2
global outvolts
print "Learning..."
#-----------------------------------------------------
def printWeights():
global direction
global inspikes
global outspikes
global outspikes2
global outvolts
global prj
print "outspikes:", outspikes
adjust=0.02
negadjust=0.02
nmax=inspikes.index(max(inspikes))
if (outspikes[0]<outspikes[1]) and (direction==dirs[1]):
print 'correct 0'
elif (outspikes[0]>outspikes[1]) and (direction==dirs[0]):
print 'correct 3'
elif (outspikes[0]>=outspikes[1]) and (direction==dirs[1]):
print 'wrong 0'
elif (outspikes[0]<=outspikes[1]) and (direction==dirs[0]):
print 'wrong 3'
else:
print 'no 5'
print
printWeights()
#-----------------------------------------------------
for p in prj:
currentWeight = p.getWeights()[0]
print direction, ntrial, p, currentWeight
check()
def LatInhibCircuit(weights = 0.5):
inp1Num = 10
inp2Num = 10
inh1Num = 10
inh2Num = 10
inh1 = p.Population(inh1Num, cellclass=p.IF_cond_exp)
inh2 = p.Population(inh2Num, cellclass=p.IF_cond_exp)
inp1 = p.Population(inp1Num, cellclass=p.IF_cond_exp)
inp2 = p.Population(inp2Num, cellclass=p.IF_cond_exp)
p1 = p.Population(inp1Num, cellclass=p.SpikeSourcePoisson)
p2 = p.Population(inp2Num, cellclass=p.SpikeSourcePoisson)
rate1 = 300./inp1Num
rate2 = 600./inp2Num
p1.set('rate', rate1)
p2.set('rate', rate2)
p1ToInp1 = p.Projection(p1, inp1, target='excitatory', method=p.OneToOneConnector())
p2ToInp2 = p.Projection(p2, inp2, target='excitatory', method=p.OneToOneConnector())
inp1toInh1 = p.Projection(inp1, inh1, target='excitatory', method=p.OneToOneConnector())
inp2toInh2 = p.Projection(inp2, inh2, target='excitatory', method=p.OneToOneConnector())
inh1ToInp2 = p.Projection(inh1, inp2, target='inhibitory', method=p.OneToOneConnector())
inh2ToInp1 = p.Projection(inh2, inp1, target='inhibitory', method=p.OneToOneConnector())
p1ToInp1.setWeights(0.05)
p2ToInp2.setWeights(0.05)
inp1toInh1.setWeights(weights)
inp2toInh2.setWeights(weights)
inh1ToInp2.setWeights(0.5)
inh2ToInp1.setWeights(0.5)
p1.record()
p2.record()
inp1.record()
inp2.record()
inh1.record()
inh2.record()
inh1.record_v()
inh2.record_v()
inp1.record_v()
inp2.record_v()
p.run(duration)
inp1V = inp1.get_v()
inp1Vtmp = [np.array(inp1V[inp1V[:,0] == 0][:,2])]
for index in range(1,inp1Num):
inp1Vtmp = np.append(inp1Vtmp, [inp1V[inp1V[:,0] == index][:,2]], axis=0)
inp1Av = np.mean(inp1Vtmp, 0)
inp2V = inp2.get_v()
inp2Vtmp = [np.array(inp2V[inp2V[:,0] == 0][:,2])]
for index in range(1,inp2Num):
inp2Vtmp = np.append(inp2Vtmp, [inp2V[inp2V[:,0] == index][:,2]], axis=0)
inp2Av = np.mean(inp2Vtmp, 0)
inh1V = inh1.get_v()
inh1Vtmp = [np.array(inh1V[inh1V[:,0] == 0][:,2])]
for index in range(1,inh1Num):
inh1Vtmp = np.append(inh1Vtmp, [inh1V[inh1V[:,0] == index][:,2]], axis=0)
inh1Av = np.mean(inh1Vtmp, 0)
inh2V = inh2.get_v()
inh2Vtmp = [np.array(inh2V[inh2V[:,0] == 0][:,2])]
for index in range(1,inh2Num):
inh2Vtmp = np.append(inh2Vtmp, [inh2V[inh2V[:,0] == index][:,2]], axis=0)
inh2Av = np.mean(inh2Vtmp, 0)
figure()
subplot(2,2,1)
plot(inp1Av)
plot(p1.getSpikes()[:,1] / timestep, [max(inp1Av)]*len(p1.getSpikes()), '|', markersize=20.0)
subplot(2,2,2)
plot(inp2Av)
plot(p2.getSpikes()[:,1] / timestep, [max(inp2Av)]*len(p2.getSpikes()), '|', markersize=20.0)
subplot(2,2,3)
plot(inh1Av)
subplot(2,2,4)
plot(inh2Av)
draw()
show()
numSpikes1 = len(inp1.getSpikes())
numSpikes2 = len(inp2.getSpikes())
rateOut1 = numSpikes1/duration
rateOut2 = numSpikes2/duration
kappaIn = (rate1-rate2)/(rate1+rate2)
kappaOut = (rateOut1-rateOut2)/(rateOut1+rateOut2)
p.reset()
return abs(kappaIn), abs(kappaOut)
