def plot(idx, params, label):
landscape = cl.__dict__[params['landscape']['mode']](
nrow, params['landscape'].get('specs', {}))
gids, ts = protocol.get_or_simulate(simulation, params)
ii = ts > recstart
clusters, sequences = seq.identify_vectors(ts[ii],
gids[ii] - 1,
nrow,
ncol,
steps=steps,
width=width,
td=td,
eps=eps)
das = []
t, x, y, c, a, s = sequences
for cid in np.unique(c):
cidx = c == cid
da = np.diff(a[cidx])
das.extend(da[~np.isnan(da)])
nclu = []
for i in range(0, 9000, 100):
q = (t >= (i + 500)) * (t < (i + 1500))
nclu.append((np.unique(c[q]).size))
nclusters.append(nclu)
ax1 = pl.subplot2grid((gridy, gridx), (0, idx))
plot_landscape_map(ax1, landscape)
ax1.set_title(label)
ax2 = pl.subplot2grid((gridy, gridx), (1, idx))
plot_rate_map(ax2, gids[ii])
plot_speed(s[~np.isnan(s)], label.split('\n')[-1])
plot_direction_changing(das, label.split('\n')[-1])
return ax1, ax2
'size': 4
}
},
{
'mode': 'Perlin_uniform',
'specs': {
'size': 4
}
},
]
simulation = 'sequence_EI_networks'
params = protocol.get_parameters(simulation).as_dict()
params.update({'landscape': landscapes[-1]})
gids, ts = protocol.get_or_simulate(simulation, params)
nrow = ncol = params['nrowE']
npop = nrow * ncol
offset = 1
idx = gids - offset < npop
gids, ts = gids[idx], ts[idx]
ts_bins = np.arange(500., 2500., 10.)
h = np.histogram2d(ts, gids - offset, bins=[ts_bins, range(npop + 1)])[0]
hh = h.reshape(-1, nrow, ncol)
fig, ax = pl.subplots(1)
a = ai.animate_image(ax, hh, vmin=0, vmax=np.max(hh))
date = datetime.datetime.now()
def plot_clustered_activity(axes, simulation, nrow, ncol, ts_bins, eps, td,
center, hanning_width):
title = ' '.join(simulation.split('_')[1:])
gids, ts = protocol.get_or_simulate(simulation)
npop = nrow * ncol
idx = ((gids - 1) < (nrow * ncol)) * (ts > wuptime)
gids, ts = gids[idx], ts[idx]
nclusters, clusters = dbscan.detect_wrap(ts, gids - 1, nrow, ncol, td, eps)
clustersize = np.bincount(clusters + 1)
ax = axes[0]
ax.clear()
steps = 10
for i in range(nclusters):
if clustersize[i] < 100:
continue
idx = (clusters == i)
ax.plot(ts[idx][::steps] / 1000., gids[idx][::steps], '.', ms=2)
ax.text(0.04,
0.95,
title,
verticalalignment='top',
horizontalalignment='left',
transform=ax.transAxes,
fontsize=8,
bbox={
'facecolor': 'white',
'pad': 3
})
yfmt = mpl.ticker.ScalarFormatter()
yfmt.set_powerlimits((0, 2))
ax.yaxis.set_major_formatter(yfmt)
ax.spines['right'].set_visible(False)
ax.spines['top'].set_visible(False)
ax.locator_params(nbins=3)
ax.set_ylabel('Neuron')
c = np.hanning(hanning_width)
cx, cy = center
g = np.unique(gids) - 1
n = 10
gx = g[((cx + n) > g // nrow) * (g // nrow > (cx - n)) *
((cy + n) > g % nrow) * (g % nrow > (cy - n))]
idx = np.in1d(gids - 1, gx)
tsc, gidsc = ts[idx], gids[idx]
h = np.histogram2d(tsc, gidsc - 1, bins=[ts_bins,
range(nrow * ncol + 1)])[0]
gidx = np.sum(h, 0) != 0
h = h[:, gidx]
hh = np.array([np.convolve(hi, c, mode='valid') for hi in h.T]).T
spike_sort = np.argsort(np.argmax(hh, 0))
hh1 = hh[:, spike_sort]
g = np.unique(gids)
np.random.shuffle(g)
idx = np.in1d(gids, g[:400])
tsc, gidsc = ts[idx], gids[idx]
h = np.histogram2d(tsc, gidsc - 1, bins=[ts_bins,
range(nrow * ncol + 1)])[0]
gidx = np.sum(h, 0) != 0
h = h[:, gidx]
hh = np.array([np.convolve(hi, c, mode='valid') for hi in h.T]).T
spike_sort = np.argsort(np.argmax(hh, 0))
hh2 = hh[:, spike_sort]
ax = axes[1]
ax.clear()
ax.matshow(hh1.T,
aspect='auto',
origin='bottom',
cmap='binary',
extent=[0, ts_bins[-1] / 1000., 0, hh1.shape[1]])
ax.yaxis.set_major_formatter(yfmt)
ax.spines['right'].set_visible(False)
ax.spines['top'].set_visible(False)
ax.xaxis.set_ticks_position('bottom')
ax.locator_params(nbins=3)
ax = axes[2]
ax.clear()
ax.matshow(hh2.T,
aspect='auto',
origin='bottom',
cmap='binary',
extent=[0, ts_bins[-1] / 1000., 0, hh1.shape[1]])
ax.yaxis.set_major_formatter(yfmt)
ax.spines['right'].set_visible(False)
ax.spines['top'].set_visible(False)
ax.xaxis.set_ticks_position('bottom')
ax.locator_params(nbins=3)
