def reduction(data, params):
# parse parameters
possible_keys = [
'n_neighbors',
'n_components',
'eigen_solver',
'tol',
'max_iter',
'path_method',
'neighbors_algorithm',
]
for item in params:
if item not in possible_keys: ERROR(item)
if isinstance(params[item], str):
exec(item + '=' + '"' + params[item] + '"')
else:
exec(item + '=' + str(params[item]))
# apply ISOMAP
X = manifold.Isomap(
n_neighbors=n_neighbors,
n_components=n_components,
eigen_solver=eigen_solver,
tol=tol,
max_iter=max_iter,
path_method=path_method,
neighbors_algorithm=neighbors_algorithm).fit_transform(data)
return X
def clustering(data, params):
# parse parameters
possible_keys = [
'n_clusters',
'affinity',
'linkage',
]
for item in params:
if item not in possible_keys: ERROR(item)
if isinstance(params[item], str):
exec(item + '=' + '"' + params[item] + '"')
else:
exec(item + '=' + str(params[item]))
# apply Agglomerative Clustering to reduced data
clusters = AgglomerativeClustering(n_clusters=n_clusters,
affinity=affinity,
linkage=linkage)
clusters.fit(data)
# Agglomerative Clustering does not give centers of clusters
# so lets try the mean of each cluster
cluster_centers = []
for i in range(n_clusters):
mask = (clusters.labels_ == i)
cluster_centers.append(mean(data[mask], axis=0))
cluster_centers = array(cluster_centers)
return [cluster_centers, clusters.labels_]
def clustering(data, params):
# parse arguments
possible_keys=['n_clusters','tol','init','n_jobs',]
for item in params:
if item not in possible_keys: ERROR(item)
if isinstance(params[item], str):
exec(item+'='+'"'+params[item]+'"')
else:
exec(item+'='+str(params[item]))
# apply KMeans to reduced data
clusters = KMeans(n_clusters=n_clusters, init=init, tol=tol)
clusters.fit(data)
return [clusters.cluster_centers_, clusters.labels_]
def clustering(data, params):
# parse arguments
possible_keys=['quantile','cluster_all',]
for item in params:
if item not in possible_keys: ERROR(item)
if isinstance(params[item], str):
exec(item+'='+'"'+params[item]+'"')
else:
exec(item+'='+str(params[item]))
# apply Mean Shift to reduced data
bandwidth = estimate_bandwidth(data, quantile=quantile)
clusters = MeanShift(bandwidth, cluster_all=cluster_all)
clusters.fit(data)
return [clusters.cluster_centers_, clusters.labels_]
def reduction(data, params):
# parse parameters
possible_keys = [
'components',
]
for item in params:
if item not in possible_keys: ERROR(item)
if isinstance(params[item], str):
exec(item + '=' + '"' + params[item] + '"')
else:
exec(item + '=' + str(params[item]))
# apply PCA
pca = PCA(n_components=n_components)
pca.fit(data)
X = pca.transform(data)
return X
def reduction(data, params):
# parse arguments
possible_keys = [
'data_errors_file',
'n_components',
'smooth',
'n_iter',
]
for item in params:
if item not in possible_keys: ERROR(item)
if isinstance(params[item], str):
exec(item + '=' + '"' + params[item] + '"')
else:
exec(item + '=' + str(params[item]))
# read errors file name
# data_errors_file = '../data/errors.dat'
if errors_file == None:
weights = None
else:
errors = loadtxt(errors_file)
weights = 1. / (errors)**2
# apply EMPCA
centered_der = data - mean(data, 0)
# m = empca(centered_der, 1./(errors)**2, nvec=5, smooth=0, niter=50)
m = empca(centered_der,
weights,
nvec=n_components,
smooth=smooth,
niter=n_iter,
silent=True)
X = m.coeff
return X
def clustering(data, params):
# parse parameters
possible_keys = [
'eps',
'min_samples',
'metric',
'algorithm',
'leaf_size',
]
for item in params:
if item not in possible_keys: ERROR(item)
if isinstance(params[item], str):
exec(item + '=' + '"' + params[item] + '"')
else:
exec(item + '=' + str(params[item]))
# apply DBSCAN to reduced data
clusters = DBSCAN(eps=eps,
min_samples=min_samples,
metric=metric,
algorithm=algorithm,
leaf_size=leaf_size).fit(data)
labels = clusters.labels_
n_clusters = len(set(labels)) - (1 if -1 in labels else 0)
# DBSCAN does not give centers of clusters
# so lets try the mean of each cluster
cluster_centers = []
for i in range(n_clusters):
mask = (clusters.labels_ == i)
cluster_centers.append(mean(data[mask], axis=0))
cluster_centers = array(cluster_centers)
return [cluster_centers, clusters.labels_]
from rpy2.robjects import r
import numpy as np
import os,sys
from aux import ERROR
def reduction(data, params):
in_param = ''
# parse parameters
possible_keys=['n_layers','training_frame','activation','autoencoder','hidden','epochs','ignore_const_cols',]
for item in params:
if item not in possible_keys: ERROR(item)
if isinstance(params[item], str):
in_param += ', '+item+' = '+params[item]
else:
exec(item+'='+str(params[item]))
# apply PCA
size_data = np.shape(data)[1]
in_param = 'x=1:%d'%(size_data) + ''.join([ ', '+item+' = '+params[item] for item in params if isinstance(params[item], str)])
# print('')
# print(in_param[9:])
# print(in_param)
# print('\n')
current_path = os.path.abspath('')