def test_nans(self):
# Spikes in 1/4 of the arena, uniform occupancy
rate_map = np.array([10., 0., 0., np.nan])
px = np.array([.25, .25, .25, .25])
rate = information_rate(rate_map, px)
spec = information_specificity(rate_map, px)
assert rate == 5 # bits/s
assert spec == 2 # bits/spike
def test_basic(self):
# Spikes in half of the arena, uniform occupancy
rate_map = np.array([10., 0.])
px = np.array([.5, .5])
rate = information_rate(rate_map, px)
spec = information_specificity(rate_map, px)
assert rate == 5 # bits/s
assert spec == 1 # bits/spike
rate_map = np.array([
[10., 0.],
[10., 0.],
])
px = np.array([
[.25, .25],
[.25, .25],
])
rate = information_rate(rate_map, px)
spec = information_specificity(rate_map, px)
assert rate == 5 # bits/s
assert spec == 1 # bits/spike
# Spikes in 1/4 of the arena, uniform occupancy
rate_map = np.array([10., 0., 0., 0.])
px = np.array([.25, .25, .25, .25])
rate = information_rate(rate_map, px)
spec = information_specificity(rate_map, px)
assert rate == 5 # bits/s
assert spec == 2 # bits/spike
rate_map = np.array([
[10., 0.],
[ 0., 0.],
])
px = np.array([
[.25, .25],
[.25, .25],
])
rate = information_rate(rate_map, px)
spec = information_specificity(rate_map, px)
assert rate == 5 # bits/s
assert spec == 2 # bits/spike
def visitDictDataSet(self, ds, **kw):
data = ds.data
if "analysis" not in data.keys():
data["analysis"] = {}
if "i_fields" not in data["analysis"].keys():
data["analysis"]["i_fields"] = {}
outputRoot = data["analysis"]["i_fields"]
simT = self.getOption(data, "time") # ms
jobNum = self.getOption(data, "job_num")
trialNum = kw.get("trialNum", 0)
self.createOutputDirs()
fileNameTemplate = "{0}/{1}/job{2:05}_trial{3:03}".format(self.rootDir, self.outputDir, jobNum, trialNum)
pos_x = self.getNetParam(data, "rat_pos_x")
pos_y = self.getNetParam(data, "rat_pos_y")
rat_dt = self.getNetParam(data, "rat_dt")
velocityStart = self.getOption(data, "theta_start_t")
monName_i = "spikeMon_i"
gridSep = self.getOption(data, "gridSep")
corr_cutRmin = gridSep / 2
spikes_i = DictDSVisitor._getNeuronSpikeTrain(data, monName_i, self.neuronNum)
spikes_i = self._shiftSpikeTimes(spikes_i, velocityStart)
out = {}
if self.po.bump:
if not self._dataPresent(outputRoot, "bump_i"):
if self.bumpTStart is None:
bumpTStart = velocityStart
if self.bumpTEnd is None:
bumpTEnd = bumpTStart + 1e3
senders_i, times_i, N_i = DictDSVisitor._getSpikeTrain(self, data, monName_i, ["Ne_x", "Ne_y"])
sp_i = PopulationSpikes(N_i, senders_i, times_i)
Fe = sp_i.avgFiringRate(bumpTStart, bumpTEnd)
Ne_x = self.getNetParam(data, "Ne_x")
Ne_y = self.getNetParam(data, "Ne_y")
bump_i = np.reshape(Fe, (Ne_y, Ne_x))
out["bump_i"] = bump_i
else:
log_info("IGridPlotVisitor", "Bump data present. Skipping analysis.")
if self.po.spikes:
if not self._dataPresent(outputRoot, "spikes_i"):
out["spikes_i"] = spikes_i
else:
log_info("IGridPlotVisitor", "Spike data present. Skipping analysis.")
if self.po.rateMap:
if not self._dataPresent(
outputRoot, "rateMap_i", "rateMap_i_X", "rateMap_i_Y", "info_specificity", "sparsity"
):
figure()
# This should speed up info/sparsity computation
if not self._dataPresent(outputRoot, "rateMap_i", "rateMap_i_X", "rateMap_i_Y"):
rateMap_i, xedges_i, yedges_i = SNSpatialRate2D(
spikes_i, pos_x, pos_y, rat_dt, self.arenaDiam, self.smoothingSigma
)
rateMap_i *= 1e3 # should be Hz
X, Y = np.meshgrid(xedges_i, yedges_i)
else:
rateMap_i = outputRoot["rateMap_i"]
X = outputRoot["rateMap_i_X"]
Y = outputRoot["rateMap_i_Y"]
px = occupancy_prob_dist(spikes_i, pos_x, pos_y, rat_dt, self.arenaDiam, self.smoothingSigma)
info_i = information_specificity(rateMap_i, px)
sparsity_i = spatial_sparsity(rateMap_i, px)
# pcolormesh(X, Y, rateMap_i)
# colorbar()
# axis('equal')
# axis('off')
# savefig('{0}_rateMap_I.png'.format(fileNameTemplate))
out["rateMap_i"] = rateMap_i
out["rateMap_i_X"] = X
out["rateMap_i_Y"] = Y
out["info_specificity"] = info_i
out["sparsity"] = sparsity_i
else:
log_info("IGridPlotVisitor", "Rate map data present. Skipping analysis.")
if self.po.fft:
if not self._dataPresent(outputRoot, "FFTX", "FFTY", "FFT"):
FT_size = 256
FX_i, FY_i, PSD_i_centered = self.computeFFT(rateMap_i, FT_size)
out["FFTX"] = FX_i
out["FFTY"] = FY_i
out["FFT"] = PSD_i_centered
else:
log_info("IGridPlotVisitor", "FFT data present. Skipping analysis.")
if self.po.sn_ac:
if not self._dataPresent(outputRoot, "corr_X", "corr_Y", "corr_i"):
corr_i, xedges_corr_i, yedges_corr_i = SNAutoCorr(rateMap_i, self.arenaDiam, self.smoothingSigma)
X, Y = np.meshgrid(xedges_corr_i, yedges_corr_i)
out["corr_X"] = X
out["corr_Y"] = Y
out["corr_i"] = corr_i
else:
log_info("IGridPlotVisitor", "Single neuron AC data present. Skipping analysis.")
if self.po.gridness_ac:
if not self._dataPresent(outputRoot, "gridnessScore", "gridnessCorr", "gridnessAngles"):
G_i, crossCorr_i, angles_i = cellGridnessScore(
rateMap_i, self.arenaDiam, self.smoothingSigma, corr_cutRmin
)
# Gridness score valid only when T >= minGridnessT
spikeTimes = data["spikeMon_i"]["events"]["times"]
lastSpikeT = spikeTimes[-1] if len(spikeTimes) != 0 else np.nan
if lastSpikeT >= self.minGridnessT:
out["gridnessScore"] = G_i
else:
log_warn("IGridPlotVisitor", "Simulation too short, G_i <- NaN")
out["gridnessScore"] = np.nan
out["gridnessCorr"] = crossCorr_i
out["gridnessAngles"] = angles_i
else:
log_info("IGridPlotVisitor", "Gridness AC data present. Skipping analysis.")
data["analysis"]["i_fields"].update(out)
def visitDictDataSet(self, ds, **kw):
data = ds.data
if "analysis" not in data.keys():
data["analysis"] = {}
outputRoot = data["analysis"]
simT = self.getOption(data, "time") # ms
jobNum = self.getOption(data, "job_num")
trialNum = kw.get("trialNum", 0)
self.createOutputDirs()
fileNameTemplate = "{0}/{1}/job{2:05}_trial{3:03}".format(self.rootDir, self.outputDir, jobNum, trialNum)
pos_x = self.getNetParam(data, "rat_pos_x")
pos_y = self.getNetParam(data, "rat_pos_y")
rat_dt = self.getNetParam(data, "rat_dt")
velocityStart = self.getOption(data, "theta_start_t")
monName_e = "spikeMon_e"
gridSep = self.getOption(data, "gridSep")
corr_cutRmin = gridSep / 2
spikes_e = DictDSVisitor._getNeuronSpikeTrain(data, monName_e, self.neuronNum)
spikes_e = self._shiftSpikeTimes(spikes_e, velocityStart)
out = {}
if self.po.bump:
if not self._dataPresent(outputRoot, "bump_e"):
figure()
if self.bumpTStart is None:
bumpTStart = velocityStart
if self.bumpTEnd is None:
bumpTEnd = bumpTStart + 1e3
# E population bump attractor
senders_e, times_e, N_e = DictDSVisitor._getSpikeTrain(self, data, monName_e, ["Ne_x", "Ne_y"])
sp_e = PopulationSpikes(N_e, senders_e, times_e)
Fe = sp_e.avgFiringRate(bumpTStart, bumpTEnd)
Ne_x = self.getNetParam(data, "Ne_x")
Ne_y = self.getNetParam(data, "Ne_y")
bump_e = np.reshape(Fe, (Ne_y, Ne_x))
torusFiringRate(rateMap=bump_e, labelx="", labely="Neuron #", titleStr="E firing rate")
savefig(fileNameTemplate + "_bump_E.png")
out["bump_e"] = bump_e
else:
log_info("GridPlotVisitor", "Bump data present. Skipping analysis.")
if self.po.spikes:
if not self._dataPresent(outputRoot, "spikes_e", "rat_pos_x", "rat_pos_y", "rat_dt"):
# E cell spikes
figure()
plotSpikes2D(spikes_e, pos_x, pos_y, rat_dt)
savefig(fileNameTemplate + "_spikePlot_E.png")
out["spikes_e"] = spikes_e
out["rat_pos_x"] = pos_x
out["rat_pos_y"] = pos_y
out["rat_dt"] = rat_dt
else:
log_info("GridPlotVisitor", "Spike data present. Skipping analysis.")
if self.po.rateMap:
if not self._dataPresent(
outputRoot, "rateMap_e", "rateMap_e_X", "rateMap_e_Y", "info_specificity", "sparsity"
):
# E cell rate map
figure()
# This should speed up info/sparsity computation
if not self._dataPresent(outputRoot, "rateMap_e", "rateMap_e_X", "rateMap_e_Y"):
rateMap_e, xedges_e, yedges_e = SNSpatialRate2D(
spikes_e, pos_x, pos_y, rat_dt, self.arenaDiam, self.smoothingSigma
)
rateMap_e *= 1e3 # should be Hz
X, Y = np.meshgrid(xedges_e, yedges_e)
else:
rateMap_e = outputRoot["rateMap_e"]
X = outputRoot["rateMap_e_X"]
Y = outputRoot["rateMap_e_Y"]
px = occupancy_prob_dist(spikes_e, pos_x, pos_y, rat_dt, self.arenaDiam, self.smoothingSigma)
info_e = information_specificity(rateMap_e, px)
sparsity_e = spatial_sparsity(rateMap_e, px)
pcolormesh(X, Y, rateMap_e)
colorbar()
axis("equal")
axis("off")
savefig("{0}_rateMap_E.png".format(fileNameTemplate))
out["rateMap_e"] = rateMap_e
out["rateMap_e_X"] = X
out["rateMap_e_Y"] = Y
out["info_specificity"] = info_e
out["sparsity"] = sparsity_e
else:
log_info("GridPlotVisitor", "Rate map data present. Skipping analysis.")
if self.po.fft:
if not self._dataPresent(outputRoot, "FFTX", "FFTY", "FFT"):
# E cell rate map FFT
figure()
FT_size = 256
FX_e, FY_e, PSD_e_centered = self.computeFFT(rateMap_e, FT_size)
pcolormesh(FX_e, FY_e, PSD_e_centered)
# axis('equal')
xlim([-10, 10])
ylim([-10, 10])
savefig("{0}_fft2_E.png".format(fileNameTemplate))
out["FFTX"] = FX_e
out["FFTY"] = FY_e
out["FFT"] = PSD_e_centered
else:
log_info("GridPlotVisitor", "FFT data present. Skipping analysis.")
if self.po.sn_ac:
if not self._dataPresent(outputRoot, "corr_X", "corr_Y", "corr"):
# E cell autocorrelation
figure()
corr_e, xedges_corr_e, yedges_corr_e = SNAutoCorr(rateMap_e, self.arenaDiam, self.smoothingSigma)
X, Y = np.meshgrid(xedges_corr_e, yedges_corr_e)
pcolormesh(X, Y, corr_e)
axis("equal")
axis("off")
savefig("{0}_rateCorr_E.png".format(fileNameTemplate))
out["corr_X"] = X
out["corr_Y"] = Y
out["corr"] = corr_e
else:
log_info("GridPlotVisitor", "Single neuron AC data present. Skipping analysis.")
if self.po.gridness_ac:
if not self._dataPresent(outputRoot, "gridnessScore", "gridnessCorr", "gridnessAngles"):
# E cell gridness correlations
figure()
G_e, crossCorr_e, angles_e = cellGridnessScore(
rateMap_e, self.arenaDiam, self.smoothingSigma, corr_cutRmin
)
plot(angles_e, crossCorr_e)
xlabel("Angle (deg.)")
ylabel("Corr. coefficient")
savefig("{0}_gridnessCorr_E.png".format(fileNameTemplate))
# Gridness score valid only when T >= minGridnessT
spikeTimes = data["spikeMon_e"]["events"]["times"]
lastSpikeT = spikeTimes[-1] if len(spikeTimes) != 0 else np.nan
if lastSpikeT >= self.minGridnessT:
out["gridnessScore"] = G_e
else:
log_warn("GridPlotVisitor", "Simulation too short, G_e <- NaN")
out["gridnessScore"] = np.nan
out["gridnessCorr"] = crossCorr_e
out["gridnessAngles"] = angles_e
else:
log_info("GridPlotVisitor", "Gridness AC data present. Skipping analysis.")
data["analysis"].update(out)
def visitDictDataSet(self, ds, **kw):
data = ds.data
if 'analysis' not in data.keys():
data['analysis'] = {}
if 'i_fields' not in data['analysis'].keys():
data['analysis']['i_fields'] = {}
outputRoot = data['analysis']['i_fields']
simT = self.getOption(data, 'time') # ms
jobNum = self.getOption(data, 'job_num')
trialNum = kw.get('trialNum', 0)
self.createOutputDirs()
fileNameTemplate = "{0}/{1}/job{2:05}_trial{3:03}".format(
self.rootDir, self.outputDir, jobNum, trialNum)
pos_x = self.getNetParam(data, 'rat_pos_x')
pos_y = self.getNetParam(data, 'rat_pos_y')
rat_dt = self.getNetParam(data, 'rat_dt')
velocityStart = self.getOption(data, 'theta_start_t')
monName_i = 'spikeMon_i'
gridSep = self.getOption(data, 'gridSep')
corr_cutRmin = gridSep / 2
spikes_i = DictDSVisitor._getNeuronSpikeTrain(data, monName_i,
self.neuronNum)
spikes_i = self._shiftSpikeTimes(spikes_i, velocityStart)
out = {}
if self.po.bump:
if not self._dataPresent(outputRoot, 'bump_i'):
if self.bumpTStart is None:
bumpTStart = velocityStart
if self.bumpTEnd is None:
bumpTEnd = bumpTStart + 1e3
senders_i, times_i, N_i = DictDSVisitor._getSpikeTrain(
self, data, monName_i, ['Ne_x', 'Ne_y'])
sp_i = PopulationSpikes(N_i, senders_i, times_i)
Fe = sp_i.avgFiringRate(bumpTStart, bumpTEnd)
Ne_x = self.getNetParam(data, 'Ne_x')
Ne_y = self.getNetParam(data, 'Ne_y')
bump_i = np.reshape(Fe, (Ne_y, Ne_x))
out['bump_i'] = bump_i
else:
log_info("IGridPlotVisitor",
"Bump data present. Skipping analysis.")
if self.po.spikes:
if not self._dataPresent(outputRoot, 'spikes_i'):
out['spikes_i'] = spikes_i
else:
log_info("IGridPlotVisitor",
"Spike data present. Skipping analysis.")
if self.po.rateMap:
if not self._dataPresent(outputRoot, 'rateMap_i', 'rateMap_i_X',
'rateMap_i_Y', 'info_specificity',
'sparsity'):
figure()
# This should speed up info/sparsity computation
if not self._dataPresent(outputRoot, 'rateMap_i',
'rateMap_i_X', 'rateMap_i_Y'):
rateMap_i, xedges_i, yedges_i = SNSpatialRate2D(
spikes_i, pos_x, pos_y, rat_dt, self.arenaDiam,
self.smoothingSigma)
rateMap_i *= 1e3 # should be Hz
X, Y = np.meshgrid(xedges_i, yedges_i)
else:
rateMap_i = outputRoot['rateMap_i']
X = outputRoot['rateMap_i_X']
Y = outputRoot['rateMap_i_Y']
px = occupancy_prob_dist(spikes_i, pos_x, pos_y, rat_dt,
self.arenaDiam, self.smoothingSigma)
info_i = information_specificity(rateMap_i, px)
sparsity_i = spatial_sparsity(rateMap_i, px)
#pcolormesh(X, Y, rateMap_i)
#colorbar()
#axis('equal')
#axis('off')
#savefig('{0}_rateMap_I.png'.format(fileNameTemplate))
out['rateMap_i'] = rateMap_i
out['rateMap_i_X'] = X
out['rateMap_i_Y'] = Y
out['info_specificity'] = info_i
out['sparsity'] = sparsity_i
else:
log_info("IGridPlotVisitor",
"Rate map data present. Skipping analysis.")
if self.po.fft:
if not self._dataPresent(outputRoot, 'FFTX', 'FFTY', 'FFT'):
FT_size = 256
FX_i, FY_i, PSD_i_centered = self.computeFFT(
rateMap_i, FT_size)
out['FFTX'] = FX_i
out['FFTY'] = FY_i
out['FFT'] = PSD_i_centered
else:
log_info("IGridPlotVisitor",
"FFT data present. Skipping analysis.")
if self.po.sn_ac:
if not self._dataPresent(outputRoot, 'corr_X', 'corr_Y', 'corr_i'):
corr_i, xedges_corr_i, yedges_corr_i = SNAutoCorr(
rateMap_i, self.arenaDiam, self.smoothingSigma)
X, Y = np.meshgrid(xedges_corr_i, yedges_corr_i)
out['corr_X'] = X
out['corr_Y'] = Y
out['corr_i'] = corr_i
else:
log_info("IGridPlotVisitor",
"Single neuron AC data present. Skipping analysis.")
if self.po.gridness_ac:
if not self._dataPresent(outputRoot, 'gridnessScore',
'gridnessCorr', 'gridnessAngles'):
G_i, crossCorr_i, angles_i = cellGridnessScore(
rateMap_i, self.arenaDiam, self.smoothingSigma,
corr_cutRmin)
# Gridness score valid only when T >= minGridnessT
spikeTimes = data['spikeMon_i']['events']['times']
lastSpikeT = spikeTimes[-1] if len(spikeTimes) != 0 else np.nan
if lastSpikeT >= self.minGridnessT:
out['gridnessScore'] = G_i
else:
log_warn('IGridPlotVisitor',
'Simulation too short, G_i <- NaN')
out['gridnessScore'] = np.nan
out['gridnessCorr'] = crossCorr_i
out['gridnessAngles'] = angles_i
else:
log_info("IGridPlotVisitor",
"Gridness AC data present. Skipping analysis.")
data['analysis']['i_fields'].update(out)
def visitDictDataSet(self, ds, **kw):
data = ds.data
if 'analysis' not in data.keys():
data['analysis'] = {}
outputRoot = data['analysis']
simT = self.getOption(data, 'time') # ms
jobNum = self.getOption(data, 'job_num')
trialNum = kw.get('trialNum', 0)
self.createOutputDirs()
fileNameTemplate = "{0}/{1}/job{2:05}_trial{3:03}".format(
self.rootDir, self.outputDir, jobNum, trialNum)
pos_x = self.getNetParam(data, 'rat_pos_x')
pos_y = self.getNetParam(data, 'rat_pos_y')
rat_dt = self.getNetParam(data, 'rat_dt')
velocityStart = self.getOption(data, 'theta_start_t')
monName_e = 'spikeMon_e'
gridSep = self.getOption(data, 'gridSep')
corr_cutRmin = gridSep / 2
spikes_e = DictDSVisitor._getNeuronSpikeTrain(data, monName_e,
self.neuronNum)
spikes_e = self._shiftSpikeTimes(spikes_e, velocityStart)
out = {}
if self.po.bump:
if not self._dataPresent(outputRoot, 'bump_e'):
figure()
if self.bumpTStart is None:
bumpTStart = velocityStart
if self.bumpTEnd is None:
bumpTEnd = bumpTStart + 1e3
# E population bump attractor
senders_e, times_e, N_e = DictDSVisitor._getSpikeTrain(
self, data, monName_e, ['Ne_x', 'Ne_y'])
sp_e = PopulationSpikes(N_e, senders_e, times_e)
Fe = sp_e.avgFiringRate(bumpTStart, bumpTEnd)
Ne_x = self.getNetParam(data, 'Ne_x')
Ne_y = self.getNetParam(data, 'Ne_y')
bump_e = np.reshape(Fe, (Ne_y, Ne_x))
torusFiringRate(rateMap=bump_e,
labelx='',
labely='Neuron #',
titleStr='E firing rate')
savefig(fileNameTemplate + '_bump_E.png')
out['bump_e'] = bump_e
else:
log_info("GridPlotVisitor",
"Bump data present. Skipping analysis.")
if self.po.spikes:
if not self._dataPresent(outputRoot, 'spikes_e', 'rat_pos_x',
'rat_pos_y', 'rat_dt'):
# E cell spikes
figure()
plotSpikes2D(spikes_e, pos_x, pos_y, rat_dt)
savefig(fileNameTemplate + '_spikePlot_E.png')
out['spikes_e'] = spikes_e
out['rat_pos_x'] = pos_x
out['rat_pos_y'] = pos_y
out['rat_dt'] = rat_dt
else:
log_info("GridPlotVisitor",
"Spike data present. Skipping analysis.")
if self.po.rateMap:
if not self._dataPresent(outputRoot, 'rateMap_e', 'rateMap_e_X',
'rateMap_e_Y', 'info_specificity',
'sparsity'):
# E cell rate map
figure()
# This should speed up info/sparsity computation
if not self._dataPresent(outputRoot, 'rateMap_e',
'rateMap_e_X', 'rateMap_e_Y'):
rateMap_e, xedges_e, yedges_e = SNSpatialRate2D(
spikes_e, pos_x, pos_y, rat_dt, self.arenaDiam,
self.smoothingSigma)
rateMap_e *= 1e3 # should be Hz
X, Y = np.meshgrid(xedges_e, yedges_e)
else:
rateMap_e = outputRoot['rateMap_e']
X = outputRoot['rateMap_e_X']
Y = outputRoot['rateMap_e_Y']
px = occupancy_prob_dist(spikes_e, pos_x, pos_y, rat_dt,
self.arenaDiam, self.smoothingSigma)
info_e = information_specificity(rateMap_e, px)
sparsity_e = spatial_sparsity(rateMap_e, px)
pcolormesh(X, Y, rateMap_e)
colorbar()
axis('equal')
axis('off')
savefig('{0}_rateMap_E.png'.format(fileNameTemplate))
out['rateMap_e'] = rateMap_e
out['rateMap_e_X'] = X
out['rateMap_e_Y'] = Y
out['info_specificity'] = info_e
out['sparsity'] = sparsity_e
else:
log_info("GridPlotVisitor",
"Rate map data present. Skipping analysis.")
if self.po.fft:
if not self._dataPresent(outputRoot, 'FFTX', 'FFTY', 'FFT'):
# E cell rate map FFT
figure()
FT_size = 256
FX_e, FY_e, PSD_e_centered = self.computeFFT(
rateMap_e, FT_size)
pcolormesh(FX_e, FY_e, PSD_e_centered)
#axis('equal')
xlim([-10, 10])
ylim([-10, 10])
savefig('{0}_fft2_E.png'.format(fileNameTemplate))
out['FFTX'] = FX_e
out['FFTY'] = FY_e
out['FFT'] = PSD_e_centered
else:
log_info("GridPlotVisitor",
"FFT data present. Skipping analysis.")
if self.po.sn_ac:
if not self._dataPresent(outputRoot, 'corr_X', 'corr_Y', 'corr'):
# E cell autocorrelation
figure()
corr_e, xedges_corr_e, yedges_corr_e = SNAutoCorr(
rateMap_e, self.arenaDiam, self.smoothingSigma)
X, Y = np.meshgrid(xedges_corr_e, yedges_corr_e)
pcolormesh(X, Y, corr_e)
axis('equal')
axis('off')
savefig('{0}_rateCorr_E.png'.format(fileNameTemplate))
out['corr_X'] = X
out['corr_Y'] = Y
out['corr'] = corr_e
else:
log_info("GridPlotVisitor",
"Single neuron AC data present. Skipping analysis.")
if self.po.gridness_ac:
if not self._dataPresent(outputRoot, 'gridnessScore',
'gridnessCorr', 'gridnessAngles'):
# E cell gridness correlations
figure()
G_e, crossCorr_e, angles_e = cellGridnessScore(
rateMap_e, self.arenaDiam, self.smoothingSigma,
corr_cutRmin)
plot(angles_e, crossCorr_e)
xlabel('Angle (deg.)')
ylabel('Corr. coefficient')
savefig('{0}_gridnessCorr_E.png'.format(fileNameTemplate))
# Gridness score valid only when T >= minGridnessT
spikeTimes = data['spikeMon_e']['events']['times']
lastSpikeT = spikeTimes[-1] if len(spikeTimes) != 0 else np.nan
if lastSpikeT >= self.minGridnessT:
out['gridnessScore'] = G_e
else:
log_warn('GridPlotVisitor',
'Simulation too short, G_e <- NaN')
out['gridnessScore'] = np.nan
out['gridnessCorr'] = crossCorr_e
out['gridnessAngles'] = angles_e
else:
log_info("GridPlotVisitor",
"Gridness AC data present. Skipping analysis.")
data['analysis'].update(out)
def visitDictDataSet(self, ds, **kw):
data = ds.data
n_recorded = 510
Nneurons = 100
if 'analysis' not in data.keys():
data['analysis'] = {}
if 'i_fields' not in data['analysis'].keys():
data['analysis']['i_fields'] = {}
data['analysis']['i_fields']['neurons'] = []
for neuron_num in np.random.choice(n_recorded, Nneurons, replace=False):
data['analysis']['i_fields']['neurons'].append({})
outputRoot = data['analysis']['i_fields']['neurons'][-1]
simT = self.getOption(data, 'time') # ms
jobNum = self.getOption(data, 'job_num')
trialNum = kw.get('trialNum', 0)
self.createOutputDirs()
fileNameTemplate = "{0}/{1}/job{2:05}_trial{3:03}_nrn{4:02}".format(self.rootDir,
self.outputDir,
jobNum, trialNum,
neuron_num)
pos_x = self.getNetParam(data, 'rat_pos_x')
pos_y = self.getNetParam(data, 'rat_pos_y')
rat_dt = self.getNetParam(data, 'rat_dt')
velocityStart = self.getOption(data, 'theta_start_t')
monName_i = 'spikeMon_i'
gridSep = self.getOption(data, 'gridSep')
corr_cutRmin = gridSep / 2
spikes_i = DictDSVisitor._getNeuronSpikeTrain(data, monName_i, neuron_num)
spikes_i = self._shiftSpikeTimes(spikes_i, velocityStart)
out = {}
if self.po.bump:
if not self._dataPresent(outputRoot, 'bump_i'):
if self.bumpTStart is None:
bumpTStart = velocityStart
if self.bumpTEnd is None:
bumpTEnd = bumpTStart + 1e3
senders_i, times_i, N_i = DictDSVisitor._getSpikeTrain(
self, data, monName_i, ['Ne_x',
'Ne_y'])
sp_i = PopulationSpikes(N_i, senders_i, times_i)
Fe = sp_i.avgFiringRate(bumpTStart, bumpTEnd)
Ne_x = self.getNetParam(data, 'Ne_x')
Ne_y = self.getNetParam(data, 'Ne_y')
bump_i = np.reshape(Fe, (Ne_y, Ne_x))
out['bump_i'] = bump_i
else:
log_info("IGridPlotVisitor",
"Bump data present. Skipping analysis.")
if self.po.spikes:
if not self._dataPresent(outputRoot, 'spikes_i'):
out['spikes_i'] = spikes_i
else:
log_info("IGridPlotVisitor",
"Spike data present. Skipping analysis.")
if self.po.rateMap:
if not self._dataPresent(outputRoot, 'rateMap_i', 'rateMap_i_X',
'rateMap_i_Y', 'info_specificity',
'sparsity'):
figure()
# This should speed up info/sparsity computation
if not self._dataPresent(outputRoot, 'rateMap_i', 'rateMap_i_X',
'rateMap_i_Y'):
rateMap_i, xedges_i, yedges_i = SNSpatialRate2D(
spikes_i, pos_x, pos_y, rat_dt, self.arenaDiam,
self.smoothingSigma)
rateMap_i *= 1e3 # should be Hz
X, Y = np.meshgrid(xedges_i, yedges_i)
else:
rateMap_i = outputRoot['rateMap_i']
X = outputRoot['rateMap_i_X']
Y = outputRoot['rateMap_i_Y']
px = occupancy_prob_dist(spikes_i, pos_x, pos_y, rat_dt,
self.arenaDiam, self.smoothingSigma)
info_i = information_specificity(rateMap_i, px)
sparsity_i = spatial_sparsity(rateMap_i, px)
pcolormesh(X, Y, rateMap_i)
colorbar()
axis('equal')
axis('off')
savefig('{0}_rateMap_I.png'.format(fileNameTemplate))
out['rateMap_i'] = rateMap_i
out['rateMap_i_X'] = X
out['rateMap_i_Y'] = Y
out['info_specificity'] = info_i
out['sparsity'] = sparsity_i
else:
log_info("IGridPlotVisitor",
"Rate map data present. Skipping analysis.")
if self.po.fft:
if not self._dataPresent(outputRoot, 'FFTX', 'FFTY', 'FFT'):
FT_size = 256
FX_i, FY_i, PSD_i_centered = self.computeFFT(rateMap_i,
FT_size)
out['FFTX'] = FX_i
out['FFTY'] = FY_i
out['FFT'] = PSD_i_centered
else:
log_info("IGridPlotVisitor",
"FFT data present. Skipping analysis.")
if self.po.sn_ac:
if not self._dataPresent(outputRoot, 'corr_X', 'corr_Y', 'corr_i'):
corr_i, xedges_corr_i, yedges_corr_i = SNAutoCorr(
rateMap_i, self.arenaDiam, self.smoothingSigma)
X, Y = np.meshgrid(xedges_corr_i, yedges_corr_i)
out['corr_X'] = X
out['corr_Y'] = Y
out['corr_i'] = corr_i
else:
log_info("IGridPlotVisitor",
"Single neuron AC data present. Skipping analysis.")
if self.po.gridness_ac:
if not self._dataPresent(outputRoot, 'gridnessScore',
'gridnessCorr', 'gridnessAngles'):
G_i, crossCorr_i, angles_i = cellGridnessScore(
rateMap_i, self.arenaDiam, self.smoothingSigma,
corr_cutRmin)
# Gridness score valid only when T >= minGridnessT
spikeTimes = data['spikeMon_i']['events']['times']
lastSpikeT = spikeTimes[-1] if len(spikeTimes) != 0 else np.nan
if lastSpikeT >= self.minGridnessT:
out['gridnessScore'] = G_i
else:
log_warn('IGridPlotVisitor', 'Simulation too short, G_i <- NaN')
out['gridnessScore'] = np.nan
out['gridnessCorr'] = crossCorr_i
out['gridnessAngles'] = angles_i
else:
log_info("IGridPlotVisitor",
"Gridness AC data present. Skipping analysis.")
plt.close('all')
outputRoot.update(out)
def visitDictDataSet(self, ds, **kw):
data = ds.data
n_recorded = 1020
Nneurons = 100
if 'analysis' not in data.keys():
data['analysis'] = {}
data['analysis']['neurons'] = []
for neuron_num in np.random.choice(n_recorded, Nneurons, replace=False):
print('Processing neuron #', neuron_num)
data['analysis']['neurons'].append({})
outputRoot = data['analysis']['neurons'][-1]
simT = self.getOption(data, 'time') # ms
jobNum = self.getOption(data, 'job_num')
trialNum = kw.get('trialNum', 0)
self.createOutputDirs()
fileNameTemplate = "{0}/{1}/job{2:05}_trial{3:03}_nrn{4:02}".format(self.rootDir,
self.outputDir, jobNum, trialNum, neuron_num)
pos_x = self.getNetParam(data, 'rat_pos_x')
pos_y = self.getNetParam(data, 'rat_pos_y')
rat_dt = self.getNetParam(data, 'rat_dt')
velocityStart = self.getOption(data, 'theta_start_t')
monName_e = 'spikeMon_e'
gridSep = self.getOption(data, 'gridSep')
corr_cutRmin = gridSep / 2
spikes_e = DictDSVisitor._getNeuronSpikeTrain(data, monName_e, neuron_num)
spikes_e = self._shiftSpikeTimes(spikes_e, velocityStart)
out = {}
if self.po.bump:
if not self._dataPresent(outputRoot, 'bump_e'):
figure()
if self.bumpTStart is None:
bumpTStart = velocityStart
if self.bumpTEnd is None:
bumpTEnd = bumpTStart + 1e3
# E population bump attractor
senders_e, times_e, N_e = DictDSVisitor._getSpikeTrain(
self, data, monName_e, ['Ne_x',
'Ne_y'])
sp_e = PopulationSpikes(N_e, senders_e, times_e)
Fe = sp_e.avgFiringRate(bumpTStart, bumpTEnd)
Ne_x = self.getNetParam(data, 'Ne_x')
Ne_y = self.getNetParam(data, 'Ne_y')
bump_e = np.reshape(Fe, (Ne_y, Ne_x))
torusFiringRate(rateMap=bump_e,
labelx='',
labely='Neuron #',
titleStr='E firing rate')
savefig(fileNameTemplate + '_bump_E.png')
out['bump_e'] = bump_e
else:
log_info("GridPlotVisitor",
"Bump data present. Skipping analysis.")
if self.po.spikes:
if not self._dataPresent(outputRoot, 'spikes_e', 'rat_pos_x',
'rat_pos_y', 'rat_dt'):
# E cell spikes
figure()
plotSpikes2D(spikes_e, pos_x, pos_y, rat_dt)
savefig(fileNameTemplate + '_spikePlot_E.png')
out['spikes_e'] = spikes_e
out['rat_pos_x'] = pos_x
out['rat_pos_y'] = pos_y
out['rat_dt'] = rat_dt
else:
log_info("GridPlotVisitor",
"Spike data present. Skipping analysis.")
if self.po.rateMap:
if not self._dataPresent(outputRoot, 'rateMap_e', 'rateMap_e_X',
'rateMap_e_Y', 'info_specificity',
'sparsity'):
# E cell rate map
figure()
# This should speed up info/sparsity computation
if not self._dataPresent(outputRoot, 'rateMap_e', 'rateMap_e_X',
'rateMap_e_Y'):
rateMap_e, xedges_e, yedges_e = SNSpatialRate2D(
spikes_e, pos_x, pos_y, rat_dt, self.arenaDiam,
self.smoothingSigma)
rateMap_e *= 1e3 # should be Hz
X, Y = np.meshgrid(xedges_e, yedges_e)
else:
rateMap_e = outputRoot['rateMap_e']
X = outputRoot['rateMap_e_X']
Y = outputRoot['rateMap_e_Y']
px = occupancy_prob_dist(spikes_e, pos_x, pos_y, rat_dt,
self.arenaDiam, self.smoothingSigma)
info_e = information_specificity(rateMap_e, px)
sparsity_e = spatial_sparsity(rateMap_e, px)
pcolormesh(X, Y, rateMap_e)
colorbar()
axis('equal')
axis('off')
savefig('{0}_rateMap_E.png'.format(fileNameTemplate))
out['rateMap_e'] = rateMap_e
out['rateMap_e_X'] = X
out['rateMap_e_Y'] = Y
out['info_specificity'] = info_e
out['sparsity'] = sparsity_e
else:
log_info("GridPlotVisitor",
"Rate map data present. Skipping analysis.")
if self.po.fft:
if not self._dataPresent(outputRoot, 'FFTX', 'FFTY', 'FFT'):
# E cell rate map FFT
figure()
FT_size = 256
FX_e, FY_e, PSD_e_centered = self.computeFFT(rateMap_e,
FT_size)
pcolormesh(FX_e, FY_e, PSD_e_centered)
#axis('equal')
xlim([-10, 10])
ylim([-10, 10])
savefig('{0}_fft2_E.png'.format(fileNameTemplate))
out['FFTX'] = FX_e
out['FFTY'] = FY_e
out['FFT'] = PSD_e_centered
else:
log_info("GridPlotVisitor",
"FFT data present. Skipping analysis.")
if self.po.sn_ac:
if not self._dataPresent(outputRoot, 'corr_X', 'corr_Y', 'corr'):
# E cell autocorrelation
figure()
corr_e, xedges_corr_e, yedges_corr_e = SNAutoCorr(
rateMap_e, self.arenaDiam, self.smoothingSigma)
X, Y = np.meshgrid(xedges_corr_e, yedges_corr_e)
pcolormesh(X, Y, corr_e)
axis('equal')
axis('off')
savefig('{0}_rateCorr_E.png'.format(fileNameTemplate))
out['corr_X'] = X
out['corr_Y'] = Y
out['corr'] = corr_e
else:
log_info("GridPlotVisitor",
"Single neuron AC data present. Skipping analysis.")
if self.po.gridness_ac:
if not self._dataPresent(outputRoot, 'gridnessScore',
'gridnessCorr', 'gridnessAngles'):
# E cell gridness correlations
figure()
G_e, crossCorr_e, angles_e = cellGridnessScore(
rateMap_e, self.arenaDiam, self.smoothingSigma,
corr_cutRmin)
plot(angles_e, crossCorr_e)
xlabel('Angle (deg.)')
ylabel('Corr. coefficient')
savefig('{0}_gridnessCorr_E.png'.format(fileNameTemplate))
# Gridness score valid only when T >= minGridnessT
spikeTimes = data['spikeMon_e']['events']['times']
lastSpikeT = spikeTimes[-1] if len(spikeTimes) != 0 else np.nan
if lastSpikeT >= self.minGridnessT:
out['gridnessScore'] = G_e
else:
log_warn('GridPlotVisitor', 'Simulation too short, G_e <- NaN')
out['gridnessScore'] = np.nan
out['gridnessCorr'] = crossCorr_e
out['gridnessAngles'] = angles_e
else:
log_info("GridPlotVisitor",
"Gridness AC data present. Skipping analysis.")
plt.close('all')
outputRoot.update(out)
