def full_run(self):
net, input_indices, _, input_times = self.scm.build(
params=self.params,
weights=self.weights,
delay=self.delay,
terminating_neurons=self.terminating_neurons,
flag=self.flag)
res = self.sim.run(net)
output_times, output_indices = network.NetworkInstance.match_static(
input_times, input_indices, res[0]["spikes"])
analysis = network.NetworkAnalysis(input_times=input_times,
input_indices=input_indices,
output_times=output_times,
output_indices=output_indices,
data_params=self.data_params,
mat_in=self.mat_in,
mat_out=self.mat_out)
I, I_norm, fp, fn, I_start, I_norm_start, fp_start, fn_start = pyscm.scm_analysis(
analysis, res[2]["spikes"], self.delay, self.flag)
return I, I_norm, fp, fn, I_start, I_norm_start, fp_start, fn_start
def full_run(self):
net, input_indices, _, input_times = self.scm.build(params=self.params,
weights=self.weights,
delay=self.delay,
terminating_neurons=self.terminating_neurons,
flag=self.flag)
res = self.sim.run(net)
output_times, output_indices = network.NetworkInstance.match_static(
input_times,
input_indices,
res[0][
"spikes"])
analysis = network.NetworkAnalysis(input_times=input_times,
input_indices=input_indices,
output_times=output_times,
output_indices=output_indices,
data_params=self.data_params,
mat_in=self.mat_in,
mat_out=self.mat_out)
I, I_norm, fp, fn, I_start, I_norm_start, fp_start, fn_start = pyscm.scm_analysis(
analysis, res[2]["spikes"], self.delay, self.flag)
return I, I_norm, fp, fn, I_start, I_norm_start, fp_start, fn_start
# Plot
for pIdx, pop in enumerate(res):
if (not "spikes" in pop):
continue
output_times = pop["spikes"]
fig = plt.figure()
ax = fig.gca()
for i, spikes in enumerate(output_times):
ax.plot(spikes, [i + 1] * len(spikes), '.', color=[0, 0, 0])
ax.set_xlabel("Spike time [ms]")
ax.set_ylabel("Neuron index")
ax.set_title("Population " + str(pIdx))
# plt.show()
# Analyse
output_times, output_indices = netw.NetworkInstance.match_static(input_times,
input_indices,
res[0][
"spikes"])
analysis = netw.NetworkAnalysis(input_times=input_times,
input_indices=input_indices,
output_times=output_times,
output_indices=output_indices,
data_params=data_params,
mat_in=mat_in, mat_out=mat_out)
I, I_norm, fp, fn, I_start, I_norm_start, fp_start, fn_start = pyscm.scm_analysis(
analysis, res[2]["spikes"], delay, flag)
flag=flag)
res = sim.run(net)
# In case the max operation in analysis throws an exception, which is the
# case for unvalid runs caused by wrong weights, e.g. no spiking at all
try:
output_times, output_indices = netw.NetworkInstance.match_static(
input_times, input_indices, res[0]["spikes"])
analysis = netw.NetworkAnalysis(input_times=input_times,
input_indices=input_indices,
output_times=output_times,
output_indices=output_indices,
data_params=data_params,
mat_in=mat_in, mat_out=mat_out)
I[i], I_norm[i], fp[i], fn[i], I_start[i], I_norm_start[i], fp_start[i], \
fn_start[i] = pyscm.scm_analysis(analysis, res[2]["spikes"], delay)
if (fp[i] > data_params["n_samples"] * data_params["n_bits_out"] * 0.5):
break
# Assure that system doesn't terminate without auto-associative spiking
times = np.zeros(len(res[0]["spikes"]))
for l in xrange(len(res[0]["spikes"])):
for j in res[0]["spikes"][l]:
if (j < 199):
times[l] += 1
if (np.amax(times) <= 2):
I[i], I_norm[i], fp[i], fn[i], I_start[i], I_norm_start[i], \
fp_start[i], fn_start[i] = 0, 0, 0, 0, 0, 0, 0, 0
except:
print "except"
I[i], I_norm[i], fp[i], fn[i], I_start[i], I_norm_start[i], fp_start[i], \
res = sim.run(net)
# In case the max operation in analysis throws an exception, which is the
# case for unvalid runs caused by wrong weights, e.g. no spiking at all
try:
output_times, output_indices = netw.NetworkInstance.match_static(
input_times, input_indices, res[0]["spikes"])
analysis = netw.NetworkAnalysis(input_times=input_times,
input_indices=input_indices,
output_times=output_times,
output_indices=output_indices,
data_params=data_params,
mat_in=mat_in,
mat_out=mat_out)
I[i], I_norm[i], fp[i], fn[i], I_start[i], I_norm_start[i], fp_start[i], \
fn_start[i] = pyscm.scm_analysis(analysis, res[2]["spikes"], delay)
if (fp[i] >
data_params["n_samples"] * data_params["n_bits_out"] * 0.5):
break
# Assure that system doesn't terminate without auto-associative spiking
times = np.zeros(len(res[0]["spikes"]))
for l in xrange(len(res[0]["spikes"])):
for j in res[0]["spikes"][l]:
if (j < 199):
times[l] += 1
if (np.amax(times) <= 2):
I[i], I_norm[i], fp[i], fn[i], I_start[i], I_norm_start[i], \
fp_start[i], fn_start[i] = 0, 0, 0, 0, 0, 0, 0, 0
except:
print "except"
print "Simulation done"
# Plot
for pIdx, pop in enumerate(res):
if (not "spikes" in pop):
continue
output_times = pop["spikes"]
fig = plt.figure()
ax = fig.gca()
for i, spikes in enumerate(output_times):
ax.plot(spikes, [i + 1] * len(spikes), '.', color=[0, 0, 0])
ax.set_xlabel("Spike time [ms]")
ax.set_ylabel("Neuron index")
ax.set_title("Population " + str(pIdx))
# plt.show()
# Analyse
output_times, output_indices = netw.NetworkInstance.match_static(
input_times, input_indices, res[0]["spikes"])
analysis = netw.NetworkAnalysis(input_times=input_times,
input_indices=input_indices,
output_times=output_times,
output_indices=output_indices,
data_params=data_params,
mat_in=mat_in,
mat_out=mat_out)
I, I_norm, fp, fn, I_start, I_norm_start, fp_start, fn_start = pyscm.scm_analysis(
analysis, res[2]["spikes"], delay, flag)
