def _cluster_spike_waveforms_by_freq(self, plot_hist = False):
"""
Clusters the found spikes into simple and complex, using a gmm_nc component GMM
It uses the maximum power in the lower region of the frequency spectrum of the
spike waveforms
"""
max_powers = self._find_max_powers()[0]
gmm = GaussianMixture(self.cs_num_gmm_components, covariance_type = self.cs_cov_type, random_state=0).fit(max_powers.reshape(-1,1))
cluster_labels = gmm.predict(max_powers.reshape(-1,1))
cluster_labels = cluster_labels.reshape(max_powers.shape)
cs_indices = self.get_spike_indices()[cluster_labels == np.argmax(gmm.means_)]
if plot_hist:
plt.figure()
# uniq = np.unique(ss.d_voltage[prang] , return_counts=True)
x = np.arange(np.min(max_powers), np.max(max_powers), 1)
if self.cs_cov_type == 'tied':
gauss_mixt = np.array([p * norm.pdf(x, mu, np.sqrt(gmm.covariances_.flatten()))
for mu, p in zip(gmm.means_.flatten(), gmm.weights_)])
else:
gauss_mixt = np.array([p * norm.pdf(x, mu, sd)
for mu, sd, p in zip(gmm.means_.flatten(), np.sqrt(gmm.covariances_.flatten()), gmm.weights_)])
colors = plt.cm.jet(np.linspace(0,1,len(gauss_mixt)))
# plot histogram overlaid by gmm gaussians
for i, gmixt in enumerate(gauss_mixt):
plt.plot(x, gmixt, label = 'Gaussian '+str(i), color = colors[i])
plt.hist(max_powers.reshape(-1,1),bins=256,density=True, color='gray')
axvlines(plt.gca(), gmm.means_)
plt.show()
self.cs_indices = cs_indices
def plot_spike_peaks(self, use_filtered=False, figsize=(21, 5)):
"""
Plots the voltage signal, overlaid by spike peak times,
overlaid by lines indicating the window around the spike peaks
"""
if use_filtered:
plt.figure(figsize=figsize)
plt.plot(self.voltage_filtered)
plt.plot(self.spike_indices,
self.voltage_filtered[self.spike_indices], '.r')
axvlines(plt.gca(),
self.spike_indices +
int(np.round(self.post_window / self.dt)),
color='g')
axvlines(plt.gca(),
self.spike_indices -
int(np.round(self.pre_window / self.dt)),
color='m')
else:
plt.figure(figsize=figsize)
plt.plot(self.voltage)
plt.plot(self.spike_indices, self.voltage[self.spike_indices],
'.r')
axvlines(plt.gca(),
self.spike_indices +
int(np.round(self.post_window / self.dt)),
color='g')
axvlines(plt.gca(),
self.spike_indices -
int(np.round(self.pre_window / self.dt)),
color='m')