def run_kmeans(self):
model = KMeans(k=self.kmeans_clusters).fit(self.data)
img_labels = model.predict(self.data).pack() # For masking
sim = model.similarity(self.data)
img_sim = sim.pack()
return model, img_sim, img_labels
def run_kmeans(self):
model = KMeans(k=self.kmeans_clusters).fit(self.data)
img_labels = model.predict(self.data).pack() # For masking
sim = model.similarity(self.data)
img_sim = sim.pack()
return model, img_sim, img_labels
def run_kmeans(data,kmeans_clusters):
## Find threshold for filtering by finding the standard deviation
# std_map = data.seriesStdev().pack()
# max_map = data.seriesMax().pack()
# filtered = data.filterOnValues(lambda x: np.std(x) > np.mean(std_map[1:10,1:10,:])+0.1)
model = KMeans(k=kmeans_clusters).fit(data)
#Kmean labels
img_labels = model.predict(data).pack()
#For masking
sim = model.similarity(data)
img_sim = sim.pack()
return model, img_sim, img_labels
def run_kmeans(data, kmeans_clusters):
## Find threshold for filtering by finding the standard deviation
# std_map = data.seriesStdev().pack()
# max_map = data.seriesMax().pack()
# filtered = data.filterOnValues(lambda x: np.std(x) > np.mean(std_map[1:10,1:10,:])+0.1)
model = KMeans(k=kmeans_clusters).fit(data)
#Kmean labels
img_labels = model.predict(data).pack()
#For masking
sim = model.similarity(data)
img_sim = sim.pack()
return model, img_sim, img_labels
def run_kmeans(self):
"""
Run Kmeans based on data and number of kmean clusters given during function initialization
RETURN:
------
model : Kmeans model
img_sim : obtained using the Similarity method of thunder's KMeansModel.
Image has pixels based on how well they match the cluster they belong to.
img_labels : Predicted label for each pixel, acquired as an image of labels
"""
model = KMeans(k=self.kmeans_clusters_index).fit(self.data)
img_labels = model.predict(self.data).pack() # For masking
sim = model.similarity(self.data)
img_sim = sim.pack()
return model, img_sim, img_labels
Example standalone app for kmeans clustering
"""
import optparse
from thunder import ThunderContext, KMeans
if __name__ == "__main__":
parser = optparse.OptionParser(
description="do kmeans clustering",
usage="%prog datafile outputdir k [options]")
parser.add_option("--maxiter", type=float, default=20)
parser.add_option("--tol", type=float, default=0.001)
opts, args = parser.parse_args()
try:
datafile = args[0]
outputdir = args[1]
k = int(args[2])
except IndexError:
parser.print_usage()
raise Exception("too few arguments")
tsc = ThunderContext.start(appName="kmeans")
data = tsc.loadSeries(datafile).cache()
model = KMeans(k=k, maxIterations=opts.maxiter).fit(data)
labels = model.predict(data)
outputdir += "-kmeans"
tsc.export(model.centers, outputdir, "centers", "matlab")
tsc.export(labels, outputdir, "labels", "matlab")
nkeys=3)
data_background = tsc.loadSeries(Working_Directory + name_for_saving_files +
'.txt',
inputFormat='text',
nkeys=3)
data_background.cache()
data_filtered.center()
data_filtered.zscore()
data_filtered.cache()
std_map = data_filtered.seriesStdev().pack()
filtered = data_filtered.filterOnValues(
lambda x: np.std(x) > np.mean(std_map[1:10, 1:10, :]) + 0.1)
model = KMeans(k=10).fit(data_filtered)
#Kmean labels
img_labels = model.predict(data_filtered).pack()
#For masking
sim = model.similarity(data_filtered)
img_sim = sim.pack()
img_size_y = np.size(img_sim, 2)
img_size_x = np.size(img_sim, 1)
brainmap, unique_clrs, newclrs_rgb, newclrs_brewer, matched_pixels, kmeans_clusters_updated = make_kmeans_maps(
data_background, model.centers, img_labels, img_sim, img_size_x,
img_size_y)
#
stimulus_pulse = 1
seriesRDD = tsc.loadSeries(
'path/to/thunder/python/thunder/utils/data/fish/bin')
print seriesRDD.dims
print seriesRDD.index
normalizedRDD = seriesRDD.normalize(baseline='mean')
stddevs = (normalizedRDD.seriesStdev().values().sample(False, 0.1,
0).collect())
plt.hist(stddevs, bins=20)
plt.plot(normalizedRDD.subset(50, thresh=0.1, stat='std').T)
# perform k-means on the normalized series
ks = [5, 10, 15, 20, 30, 50, 100, 200]
models = []
for k in ks:
models.append(KMeans(k=k).fit(normalizedRDD))
# define a couple functions to score the clustering quality
def model_error_1(model):
def series_error(series):
cluster_id = model.predict(series)
center = model.centers[cluster_id]
diff = center - series
return diff.dot(diff)**0.5
return normalizedRDD.apply(series_error).sum()
def model_error_2(model):
return 1. / model.similarity(normalizedRDD).sum()
name_for_saving_files = 'All_odors_'+ filename_save_prefix+'_eachodor'
name_for_saving_figures = Working_Directory
data_filtered = tsc.loadSeries(Working_Directory+name_for_saving_files+'_filtered.txt', inputFormat='text', nkeys=3)
data_background = tsc.loadSeries(Working_Directory+name_for_saving_files+'.txt', inputFormat='text', nkeys=3)
data_background.cache()
data_filtered.center()
data_filtered.zscore()
data_filtered.cache()
std_map = data_filtered.seriesStdev().pack()
filtered = data_filtered.filterOnValues(lambda x: np.std(x) > np.mean(std_map[1:10,1:10,:])+0.1)
model = KMeans(k=10).fit(data_filtered)
#Kmean labels
img_labels = model.predict(data_filtered).pack()
#For masking
sim = model.similarity(data_filtered)
img_sim = sim.pack()
img_size_y = np.size(img_sim,2)
img_size_x = np.size(img_sim,1)
brainmap, unique_clrs, newclrs_rgb, newclrs_brewer, matched_pixels, kmeans_clusters_updated = make_kmeans_maps(data_background, model.centers, img_labels, img_sim, img_size_x, img_size_y)
#
stimulus_pulse = 1
#
plot_kmeans_maps(Working_Directory, name_for_saving_figures, name_for_saving_files, \
"""
Example standalone app for kmeans clustering
"""
import argparse
from thunder import ThunderContext, KMeans, export
if __name__ == "__main__":
parser = argparse.ArgumentParser(description="do kmeans clustering")
parser.add_argument("datafile", type=str)
parser.add_argument("outputdir", type=str)
parser.add_argument("k", type=int)
parser.add_argument("--maxiter", type=float, default=20, required=False)
parser.add_argument("--tol", type=float, default=0.001, required=False)
args = parser.parse_args()
tsc = ThunderContext.start(appName="kmeans")
data = tsc.loadSeries(args.datafile).cache()
model = KMeans(k=args.k, maxIterations=args.maxiter).fit(data)
labels = model.predict(data)
outputdir = args.outputdir + "-kmeans"
export(model.centers, outputdir, "centers", "matlab")
export(labels, outputdir, "labels", "matlab")