def test_cluster_component():
base.features.Provide("FeatureSource", DummyFeatureExtractor())
cluster_comp = components.ClusterAnalyzer("k_means", 2)
labels = cluster_comp.read_labels()
ok = (((labels[:n_spikes] == 1).all() & (labels[n_spikes:] == 2).all()) |
((labels[:n_spikes] == 2).all() & (labels[n_spikes:] == 1).all()))
ok_(ok)
def test_cluster_component_relabel():
base.features.Provide("FeatureSource", RandomFeatures())
cluster_comp = components.ClusterAnalyzer("k_means", 5)
labs = cluster_comp.labels
cluster_comp.delete_cells(1, 2, 3, 4)
cluster_comp.relabel()
labels = np.unique(cluster_comp.labels)
labels.sort()
ok_((labels == np.array([0, 1])).all())
def test_pipeline_update():
base.features.Provide("SignalSource", DummySignalSource())
base.features.Provide("SpikeMarkerSource",
components.SpikeDetector(thresh=spike_amp / 2.))
base.features.Provide("SpikeSource", components.SpikeExtractor())
base.features.Provide("FeatureSource",
components.FeatureExtractor(normalize=False))
base.features.Provide("ClusterAnalyzer",
components.ClusterAnalyzer("k_means", 2))
base.features['FeatureSource'].add_feature("P2P")
cl1 = base.features["ClusterAnalyzer"].labels
base.features["SignalSource"].update()
cl2 = base.features["ClusterAnalyzer"].labels
ok_(~(len(cl1) == len(cl2)))
def test_null_labels_returned_for_truncated_spikes():
signal_src = DummySignalSource()
signal_src._spikes = signal_src._spikes[:, :-period / 1000. * FS * 2.5]
base.features.Provide("SignalSource", signal_src)
base.features.Provide("SpikeMarkerSource", DummySpikeDetector())
base.features.Provide("SpikeSource", components.SpikeExtractor())
base.features.Provide("FeatureSource",
components.FeatureExtractor(normalize=False))
base.features.Provide("ClusterAnalyzer",
components.ClusterAnalyzer("k_means", 1))
base.features['FeatureSource'].add_feature("P2P")
cl = base.features["ClusterAnalyzer"].labels
true_labels = np.ones(n_spikes - 2)
true_labels[-1] = 0
ok_((cl == true_labels).all())
def test_cluster_component_methods_before_labels_requested():
# issue #75
# The followig methods should not fail when called before lebels are
# externally requested (cluster_labels = None)
base.features.Provide("FeatureSource", RandomFeatures())
cluster_comp = components.ClusterAnalyzer("k_means", 2)
cluster_comp.delete_cells(1)
cluster_comp.cluster_labels = None
cluster_comp.delete_spikes([0])
cluster_comp.cluster_labels = None
cluster_comp.merge_cells(1, 2)
cluster_comp.cluster_labels = None
cluster_comp.relabel()
ok_(cluster_comp.cluster_labels is not None)
def test_cluster_component_smart_update():
feature_comp = base.register("FeatureSource", DummyFeatureExtractor())
cluster_comp = components.ClusterAnalyzer("k_means", 2)
labs = cluster_comp.labels # this is a workaround to call _cluster()
# cluster_comp should NOT recluster when updated, if the number of
# spikes didn't change
cluster_comp.delete_cells(1) # modify lablels
labels_orig = cluster_comp.labels.copy()
feature_comp.add_feature('new_feature')
feature_comp.update()
labels_new_feat = cluster_comp.labels.copy()
# cluster_comp should recluster when updated, if the number of
# spikes DID change
feature_comp.add_spikes(10)
feature_comp.update()
labels_new_spikes = cluster_comp.labels.copy()
test1 = (labels_orig == labels_new_feat).all()
test2 = len(labels_new_spikes) != len(labels_new_feat)
ok_(test1 and test2)
io_filter = components.FilterStack()
base.features.Provide("RawSource", io)
base.features.Provide("EventsOutput", io)
base.features.Provide("SignalSource", io_filter)
base.features.Provide(
"SpikeMarkerSource",
MultiChannelSpikeDetector(contact=contact,
thresh=thresh,
type=type,
sp_win=[-0.6, 0.8],
resample=10,
align=True))
base.features.Provide("SpikeSource", components.SpikeExtractor(sp_win=sp_win))
base.features.Provide("FeatureSource", components.FeatureExtractor())
base.features.Provide("LabelSource",
components.ClusterAnalyzer("gmm", n_clusters))
src = base.features['SignalSource']
features = base.features['FeatureSource']
clusters = base.features['LabelSource']
events = base.features['SpikeMarkerSource']
labels = base.features['LabelSource']
browser = components.SpikeBrowserWithLabels()
plot1 = components.PlotFeaturesTimeline()
plot2 = components.PlotSpikes()
legend = components.Legend()
export = components.ExportCells()
#############################################################
# Add filters here
hd5file = os.environ['DATAPATH'] + hd5file
io_filter = components.PyTablesSource(hd5file, dataset, f_filter=filter_freq)
base.features.Provide("SignalSource", io_filter)
base.features.Provide("EventsOutput", io_filter)
base.features.Provide(
"SpikeMarkerSource",
components.SpikeDetector(contact=contact,
thresh=thresh,
type=type,
sp_win=sp_win,
resample=10,
align=True))
base.features.Provide("SpikeSource", components.SpikeExtractor(sp_win=sp_win))
base.features.Provide("FeatureSource", components.FeatureExtractor())
base.features.Provide("LabelSource", components.ClusterAnalyzer("gmm", 4))
browser = components.SpikeBrowserWithLabels()
plot1 = components.PlotFeaturesTimeline()
plot2 = components.PlotSpikes()
legend = components.Legend()
export = components.ExportCells()
#############################################################
# Add the features here:
base.features["FeatureSource"].add_feature("P2P")
base.features["FeatureSource"].add_feature("PCs", ncomps=2)
#############################################################
# Run the analysis (this can take a while)
io_filter = components.FilterStack()
base.register("RawSource", io)
base.register("EventsOutput", io)
base.register("SignalSource", io_filter)
base.register(
"SpikeMarkerSource",
components.SpikeDetector(contact=contact,
thresh=thresh,
type=detection_type,
sp_win=sp_win,
resample=1,
align=True))
base.register("SpikeSource", components.SpikeExtractor(sp_win=sp_win))
base.register("FeatureSource", components.FeatureExtractor())
base.register("LabelSource", components.ClusterAnalyzer("gmm", 4))
browser = components.SpikeBrowser()
feature_plot = components.PlotFeaturesTimeline()
wave_plot = components.PlotSpikes()
legend = components.Legend()
export = components.ExportCells()
#############################################################
# Add filters here:
base.features["SignalSource"].add_filter("LinearIIR", *filter_freq)
# Add the features here:
base.features["FeatureSource"].add_feature("P2P")
base.features["FeatureSource"].add_feature("PCA", ncomps=2)