-
Notifications
You must be signed in to change notification settings - Fork 0
/
classify.py
358 lines (258 loc) · 10.6 KB
/
classify.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
#! /usr/local/bin python
# -*- coding: iso-8859-1 -*-
# Importing packages
import numpy as np
import matplotlib.pylab as plt
from astropy.table import Table, join
#import astropy.cosmology
from sklearn.linear_model import LogisticRegression
import sys
import os
import glob
from astropy.io import ascii
from astroML.plotting import hist
from sklearn.decomposition import PCA
from sklearn import preprocessing
import machine
from sklearn.metrics import confusion_matrix
# Catalog definitions:
# 'stell' corresponds to SExtractor stellaricity
# 'Stellar_Flag' is a much better indicator that takes in account many parameters: F814W, FWHM, J-Ks, F489W-F814W.
# See Molino's paper, section "Star/galaxy separation".
def table_cleaner(table):
table.remove_column('ID')
table.remove_column('Field')
table.remove_column('Pointing')
table.remove_column('CCD')
table.remove_column('RA')
table.remove_column('Dec')
table.remove_column('x')
table.remove_column('y')
table.remove_column('photoflag')
table.remove_column('nfobs')
table.remove_column('Satur_Flag')
table.remove_column('DupliDet_Flag')
return table
create_master_catalog = False
master_catalog_name = "alhambra.Master.ALLDATA.cat"
master_catalog_directory = "data/catalogs/master_catalogs/"
# Creates the Master catalog here:
if create_master_catalog:
if not os.path.exists(master_catalog_directory):
os.makedirs(master_catalog_directory)
# Expected number of Alhambra catalogues
exp_numb_cats = 48
# NB the 6/8/14: This bug has been corrected since astropy 0.4
# Due to a bug in astropy.table (see :
# http://stackoverflow.com/questions/22617428/overflowerror-python-int-too-large-to-convert-to-c-long-with-astropy-table
# and
# https://github.com/astropy/astropy/pull/2234 )
# I truncate the first line of my catalogs until the bug is corrected.
# Truncated catalog name has a "first_col_truncated." prefix, and is in
# a different directory "/first_col_truncated/".
#catalogs_directory = 'data/catalogs/first_col_truncated/first_col_truncated.' #this line contains the repertory AND the prefix.
# When the bug is solved,use this line:
catalogs_directory = 'data/catalogs/original_catalogs/'
catalog_names = glob.glob(catalogs_directory+'alhambra.*.ColorProBPZ.cat')
print "I have found "+str(len(catalog_names))+" catalogues."
if exp_numb_cats != len(catalog_names) :
print "WARNING: The expected number of catalogues is "+str(exp_numb_cats)+". Why do we read a different number?"
i=0
for catalog_filename in catalog_names:
if catalog_filename == catalog_names[0]:
t = Table.read(catalog_filename, format='ascii')
print "first catalog"
else:
t_buffer = t
t = Table.read(catalog_filename, format='ascii')
#t = join(t_buffer, t, join_type='outer')
i=i+1
print "I have red "+ catalog_filename +" , # "+ str(i)
# Selects only the objects that are not duplicated, or, if they are, selects only the best SNR object.
t = t[:][np.where(t['DupliDet_Flag']==0)]
print "There are "+ str(len(t))+" unique elements when low SNR duplicates are deleted"
# Joins the new catalog to the master catalog being created
if catalog_filename != catalog_names[0]:
print t['Stell_Mass_1'].mean()
t = join(t_buffer, t, join_type='outer')
print t['Stel l_Mass_1'].mean()
# Writes the Master catalog:
ascii.write(t, master_catalog_directory + master_catalog_name)
np.save(master_catalog_directory + master_catalog_name + ".npy", t)
create_arrays_long = True
if create_arrays_long is True:
number_lines_to_read = 10000
t = Table.read("data/catalogs/master_catalogs/alhambra.Master.ALLDATA.cat", format='ascii', data_end=number_lines_to_read)
########
# Let's begin by looking a bit at the data
F814W_limit = 22.5 # The limit over which the Stellar_Flag is always 0.5
t_faint = t[np.where(t["F814W"]>=F814W_limit)]
t_bright = t[np.where(t["F814W"]<F814W_limit)]
if np.where(t_faint["Stellar_Flag"] != 0.5)[0].size == 0: print "Sanity check OK: all faint sources have Stellar_Flag = 0.5"
if np.where(t_bright["Stellar_Flag"] == 0.5)[0].size != 0: print "Some bright sources also have Stellar_Flag = 0.5"
# Shows that some random values of the bright sample can also have Stellar_Flag = 0.5 (which is OK)
# print t_bright[np.where(t_bright["Stellar_Flag"] == 0.5)[0]]["F814W"] # Shows teh values of F814W for these bright objects with Stellar_Flag = 0.5
plt.figure(figsize=(10,10), facecolor='w', edgecolor='k')
plt.hist(t_bright["Stellar_Flag"], bins=100)
plt.title("Bright objects Stellar_Flag distribution")
#plt.show()
plt.close()
# Gets only the objects with Stellar_Flag<0.1 or Stellar_Flag>0.9: the "sure" catalog
limit = 0.1
center = 0.5
t_bright_sure = t_bright[np.where(np.abs(t_bright["Stellar_Flag"]-center)>center-limit)]
# Sanity check:
plt.figure(figsize=(10,10), facecolor='w', edgecolor='k')
plt.hist(t_bright_sure["Stellar_Flag"], bins=100)
plt.title("Bright objects Stellar_Flag distribution, only for very certain objects")
#plt.show()
plt.close()
# Selects stars or galaxies from the "sure" catalog
t_bright_sure_star = t_bright_sure[np.where(t_bright_sure["Stellar_Flag"]>1.-limit)]
t_bright_sure_gal = t_bright_sure[np.where(t_bright_sure["Stellar_Flag"]<limit)]
# Sanity check:
plt.figure(figsize=(10,10), facecolor='w', edgecolor='k')
plt.hist(t_bright_sure_star["Stellar_Flag"], label = "star")
plt.hist(t_bright_sure_gal["Stellar_Flag"], label = "galaxy")
plt.title("Bright objects Stellar_Flag distribution for very certain stars vs galaxies")
plt.legend()
#plt.show()
plt.close()
# Plots the hists of the different variables:
make_all_plots = False
if make_all_plots:
for c in t.columns:
plt.figure(figsize=(10,10), facecolor='w', edgecolor='k')
plt.title("Distribution variable: "+c)
hist(t_bright_sure_star[c], label = "star", alpha = 0.5, bins='blocks', normed = True)
hist(t_bright_sure_gal[c], label = "gal", alpha = 0.5, bins='blocks', normed = True)
plt.legend()
plt.show()
plt.close()
t_bright_sure = table_cleaner(t_bright_sure)
X = []
for t2 in t_bright_sure:
X.append(list(t2))
X = np.array(X)
"""
#Normalization:
scaler = preprocessing.StandardScaler().fit(X)
Xnorm = scaler.transform(X)
#
# PCA
#
print "begins PCA analysis"
pca = PCA()
pca.fit(Xnorm)
X_new = pca.transform(Xnorm)
plt.figure()
plt.plot(X_new[:, 0], X_new[:, 1], 'x')
plt.xlabel('column 0')
plt.ylabel('column 1')
plt.title('column 0 x 1')
plt.figure()
plt.plot(X_new[:, 1], X_new[:, 2], 'x')
plt.xlabel('column 1')
plt.ylabel('column 2')
plt.title('column 1 x 2')
plt.figure()
plt.plot(pca.components_[0,:], 'r')
plt.plot(pca.components_[1,:], 'g')
plt.plot(pca.components_[2,:], 'b')
plt.title('loadings 0..2')
plt.show()
print(pca.explained_variance_ratio_)
sys.exit()
y_stell_sex = t["stell"]
y_stell_mol = t["Stellar_Flag"]
t.remove_column('stell')
t.remove_column('Stellar_Flag')
t.remove_column('ID')
X = []
for i in t:
X.append(list(i))
X = np.array(X)
zb_1 = t["zb_1"]
Stell_Mass_1 = t["Stell_Mass_1"]
np.save("ml/master_np_X.npy", X)
np.save("ml/master_np_y_stell_sex.npy", y_stell_sex)
np.save("ml/master_np_y_stell_mol.npy", y_stell_mol)
np.save("ml/master_zb_1.npy", zb_1)
np.save("ml/master_Stell_Mass_1.npy", Stell_Mass_1)
"""
X = np.load("ml/master_np_X.npy") # the matrix
ymol = np.load("ml/master_np_y_stell_mol.npy") # The stellaricity from Molino
ysex = np.load("ml/master_np_y_stell_sex.npy") # The stellaricity from Bertin's sextractor
zb_1 = np.load("ml/master_zb_1.npy") # z
Stell_Mass_1 = np.load("ml/master_Stell_Mass_1.npy") # stellar mass
########
# Test on the known part of the sample:
# We try only on "good" objects: Stellar_Flag>0.9 and Stellar_Flag<0.1
# half of this sample is the learning sample
# half is the target
# Target is predicted with 1% error (for a sample of 1000 "good" objects)
limit = 0.1
center = 0.5
ymol_unsure_mask = np.where(np.abs(ymol-center)>center-limit)
ymol_sure = ymol[ymol_unsure_mask]
X_sure = X[ymol_unsure_mask,:]
X_sure = X_sure[0]
ymol_bool = []
for i in ymol_sure:
if i<limit: ymol_bool.append(-1)
if i>1-limit: ymol_bool.append(1)
#plt.figure(figsize=(10,10), facecolor='w', edgecolor='k')
#plt.hist(ymol[ymol_unsure_mask])
#plt.show()
#plt.ion()
#plt.close()
# Will learn from the even positions:
X_sure_learn = X_sure[0::2]
ymol_learn = ymol_bool[0::2]
# Sill predict the odd positions
X_sure_target = X_sure[1::2]
y_target = ymol_bool[1::2]
model = machine.logistic()
model.train(X_sure_learn, ymol_learn)
y_prediction = model.use(X_sure_target)
confm = confusion_matrix(np.array(y_target), y_prediction, labels=[1, -1])
print confm[0,0]/float(np.sum(confm[0,:]))
print confm[1,1]/float(np.sum(confm[1,:]))
sys.exit()
"""
# ML
model = LogisticRegression(C=1.)
model.fit(X_sure_learn, ymol_learn)
y_prediction = model.predict(X_sure_target)
"""
# Results:
print "total number of objects to predict: "+ str(len(y_target))
N_correct = len(np.where(y_prediction - y_target == 0)[0])
print "I have predicted correctly "+str(N_correct)+" objects"
error_places = np.where(y_prediction - y_target != 0)[0]
print error_places
N_error = len(error_places)
print "there are " + str(N_error) + " errors (" +str(float(N_error)/float(N_correct)*100.)+"%)"
smsure = Stell_Mass_1[ymol_unsure_mask]
smsure_target = smsure[1::2]
print smsure_target[error_places]
Stell_Mass_1_z08 = Stell_Mass_1[np.where(np.abs(zb_1-0.8)<0.05)[0]]
tz = t[np.where(np.abs(t["zb_1"]-0.8)<0.05)[0]]
plt.hist(tz['Stell_Mass_1'], bins = 100)
plt.hist(Stell_Mass_1_z08_sure, bins = 100, alpha = 0.6)
plt.show()
sys.exit()
########
# Now we use "good" objects to predict Stellar_Flag = 0.5 objects:
model = LogisticRegression(C=1.)
model.fit(X_sure, ymol_bool) # Fit on all objects where we know the stellaricity very well (Stellar_Flag>0.9 and Stellar_Flag<0.1)
faint = np.where(ymol==0.5)[0]
print "Total number of objects: " + str(len(ymol))
print "Number of faint objects: " + str(len(faint))
X_faint = X[faint, :]
y_faint_target = model.predict(X_faint)
faint_star = np.where(y_faint_target == 1)[0]
faint_gal = np.where(y_faint_target == -1)[0]
print len(faint_star)
print len(faint_gal)
#plt.hist()