def fit_dbscan(data,
eps,
min_samples,
normalize=True,
show=True,
juxta_cluster_indices_grouped=None,
threshold_legend=None):
X = np.transpose(data)
if normalize:
from sklearn.preprocessing import minmax_scale
minmax_scale(X, feature_range=(-1, 1), axis=0, copy=False)
db = DBSCAN(eps=eps, min_samples=min_samples).fit(X)
core_samples_mask = np.zeros_like(db.labels_, dtype=bool)
core_samples_mask[db.core_sample_indices_] = True
labels = db.labels_
# Number of clusters in labels, ignoring noise if present.
n_clusters_ = len(set(labels)) - (1 if -1 in labels else 0)
score = metrics.silhouette_score(X, labels, sample_size=5000)
print('For eps={}, min_samples={}, estimated number of clusters={}'.format(
eps, min_samples, n_clusters_))
print("Silhouette Coefficient: {}".format(score))
if show:
pf.show_clustered_tsne(db, X, juxta_cluster_indices_grouped,
threshold_legend)
return db, n_clusters_, labels, core_samples_mask, score
def fit_dbscan(data,
eps,
min_samples,
show=True,
juxta_cluster_indices_grouped=None,
threshold_legend=None):
X = np.transpose(data)
db = DBSCAN(eps=eps, min_samples=min_samples).fit(X)
core_samples_mask = np.zeros_like(db.labels_, dtype=bool)
core_samples_mask[db.core_sample_indices_] = True
db_labels = db.labels_
# Number of clusters in db_labels, ignoring noise if present.
n_clusters_ = len(set(db_labels)) - (1 if -1 in db_labels else 0)
score = metrics.silhouette_score(X, db_labels, sample_size=5000)
print('For eps={}, min_samples={}, estimated number of clusters={}'.format(
eps, min_samples, n_clusters_))
print("Silhouette Coefficient: {}".format(score))
if show:
pf.show_clustered_tsne(db, X, juxta_cluster_indices_grouped,
threshold_legend)
return db, n_clusters_, db_labels, core_samples_mask, score
pf.plot_tsne(tsne,
juxta_cluster_indices_grouped,
cm=plt.cm.brg,
label_name='Peak size in uV',
label_array=(spike_thresholds_groups * 1e6).astype(int),
subtitle='T-sne of ' + str(channel_number[code]) +
'channels, ' + str(geometry_descriptions[code]))
# ----------------------------------------------------------------------------------------------------------------------
# CLUSTERING AND SCORING
db, n_clusters, labels, core_samples_mask, score = taf.fit_dbscan(
tsne, 0.019, 40, normalize=True, show=True)
pf.show_clustered_tsne(db,
tsne,
juxta_cluster_indices_grouped=None,
threshold_legend=None)
taf.calculate_precision_recall_for_single_label_grouped(
tsne,
juxta_cluster_indices_grouped,
n_clusters,
labels,
core_samples_mask,
show_means=False)
# ----------------------------------------------------------------------------------------------------------------------
# Comparing tsnes with the manual clustering of the 128 channels
base_dir = r'D:\Data\George\Projects\SpikeSorting\Joana_Paired_128ch\2015-09-03'
tsne_dir = r'Analysis\klustakwik\threshold_6_5std'
f = plt.figure(3)
ax = f.add_subplot(111)
def show_frame(ind, *nothing):
ax.clear()
ax.imshow(grayscale_resized_video_array_frame_smoothed[ind, :].reshape(
(80, 80)))
plt.figure(1)
pf.show_clustered_tsne(db,
tsne_result,
juxta_cluster_indices_grouped=None,
threshold_legend=None,
func_to_exec_on_pick=show_frame,
args_of_func_on_pick=[],
labels_on=True)
plt.figure(2)
pf.show_clustered_tsne(db,
tsne_result,
juxta_cluster_indices_grouped=None,
threshold_legend=None,
func_to_exec_on_pick=None,
args_of_func_on_pick=None,
labels_on=False)
jet = cm.jet
colors = jet(np.squeeze(preprocessing.normalize([labels], 'max')))
ax_pcs.clear()
capture_cropped.set(cv2.CAP_PROP_POS_FRAMES, int(i * frames_to_bin))
_, video_frame = capture_cropped.read()
ax_video.imshow(video_frame)
ax_pcs.imshow(reversed_pcs_video[i, :, :])
f_video = plt.figure(4)
ax_video = f_video.add_subplot(111)
f_pcs = plt.figure(5)
ax_pcs = f_pcs.add_subplot(111)
arg = [ax_video, ax_pcs]
pf.show_clustered_tsne(
dbscan_result,
tsne_video,
func_to_exec_on_pick=show_video_and_pc_reconstructed_frames,
args_of_func_on_pick=arg)
def show_video_label_on_spikes_tsne(label, ax2):
ax2.clear()
ax2.set_xlim(ax1.get_xlim())
ax2.set_ylim(ax1.get_ylim())
indices = np.squeeze(np.argwhere(video_tsne_labels == label))
ax2.scatter(tsne_spikes[indices, 0],
tsne_spikes[indices, 1],
c=[label_colors[label]])
return None
