-
Notifications
You must be signed in to change notification settings - Fork 0
/
CustomExample2.py
423 lines (369 loc) · 17.5 KB
/
CustomExample2.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
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
# Run with `python CustomExample2.py -i SpectraGAN.hdf5 --nepochs 100`
##############################################################################
# # Custom example #2
# ## Astrophysical Spectra Generation
# ### Project: Generate realistic spectra of astrophysics objects (not included)
# - Technique: generative adversarial network with a deep convolutional architecture
# - Output: generated spectra for each iteration of the neural network training (as time progresses the spectra should look more realistic) [data stored in *SpectraGAN.hdf5*]
# ### Visualization:
# - Access all the spectra via a slider that shifts through the iterations (here called "epochs")
# - Display the means and standard deviations for thos spectra
# - Calculate the means of the distributions in the browser
##############################################################################
import os
import numpy as np
import pandas as pd
import h5py
import argparse
from bokeh.models import Text
from bokeh.models import Title
from bokeh.models import Plot
from bokeh.models import Slider
from bokeh.models import Circle
from bokeh.models import Legend
from bokeh.models import Range1d
from bokeh.models import CustomJS
from bokeh.models import HoverTool
from bokeh.models import LinearAxis
from bokeh.models import ColumnDataSource
from bokeh.models import SingleIntervalTicker
from bokeh.palettes import Category10
from bokeh.plotting import figure, show, save
from bokeh.layouts import layout
parser = argparse.ArgumentParser()
parser.add_argument(
'-i',
type=str,
required=True,
help='Input file in HDF5 format.'
)
parser.add_argument(
'--nepochs',
type=int,
required=True,
help='Input file in HDF5 format.'
)
parser.add_argument(
'--no_stats',
action='store_true',
help='Consider models which do not generate means and standard deviations.'
)
FLAGS, _ = parser.parse_known_args()
# Define constants
FILENAME = FLAGS.i
RENDERERS = 'source1_init', 'source2_init'
TOTALWIDTH, WIDTHRATIO = 2500, 0.7
WIDTH, HEIGHT = int(TOTALWIDTH*WIDTHRATIO), 550
WIDTH2, HEIGHT2 = int(TOTALWIDTH*(1-WIDTHRATIO)), HEIGHT/2
NPOINTS = 3500
NEPOCHS = FLAGS.nepochs
NSTATS = 2000
NBINS = 100
NSPECTRA = 8
NDIGITS = int(np.floor(np.log10(np.abs(NEPOCHS))) + 1) #number of digits in a integer
YSHIFT = 0.15
LOGAXIS = False
# Define plotting help functions
def add_legend(fig, obj, label, **kwargs):
leg = Legend(
items=[(label, [obj])],
**kwargs)
fig.add_layout(leg, 'center')
def add_axis(fig, xlabel=None, ylabel=None, xinterval=500, yinterval=None, xbounds=None, ybounds=None):
color = '#000066'
AXIS_FORMATS = dict(
minor_tick_in=None,
minor_tick_out=None,
major_tick_in=None,
major_label_text_font_size="11pt",
major_label_text_font_style="normal",
axis_label_text_font_size="11pt",
axis_label_text_font_style="normal",
axis_line_color=color,
major_tick_line_color=color,
major_label_text_color=color,
major_tick_line_cap="round",
axis_line_cap="round",
axis_line_width=1.5,
major_tick_line_width=1.5,
)
if xlabel:
xaxis = LinearAxis(
ticker = SingleIntervalTicker(interval=xinterval),
axis_label = xlabel,
**AXIS_FORMATS
)
if xbounds:
xaxis.bounds = xbounds
fig.add_layout(xaxis, 'below')
if ylabel:
yaxis = LinearAxis(
ticker = SingleIntervalTicker(interval=yinterval),
axis_label = ylabel,
**AXIS_FORMATS
)
if ybounds:
yaxis.bounds = ybounds
fig.add_layout(yaxis, 'left')
def create_figure(w, h, title, xlims=None, ylims=None, xaxislog=False):
TOOLTIPS = [('wavelength', '$x ' + u'\u212b')]
p = figure(plot_width=w, plot_height=h, x_axis_type='log' if xaxislog else 'linear',
tools='hover,pan,wheel_zoom,box_zoom,reset', toolbar_location='above',
tooltips=TOOLTIPS)
p.toolbar.logo = None
#p.toolbar_location = None
p.axis.visible = False
p.title = Title(text=title, text_font_size='12pt')
if ylims:
p.y_range = Range1d(ylims[0],ylims[1])
if xlims:
p.x_range = Range1d(xlims[0],xlims[1])
return p
def add_graph(fig, xname, yname, renderer_source, color):
fig.circle(x=xname, y=yname, size=1, color=color, source=renderer_source)
fig.line(x=xname, y=yname, color=color, source=renderer_source)
def add_bars(fig, xname, yname, width, renderer_source, color):
fig.vbar(x=xname, top=yname, bottom=0, width=width, fill_color=color, line_color='black', source=renderer_source)
def add_vline(fig, xname, yname, width, renderer_source, color):
return fig.line(x=xname, y=yname, line_width=width, color=color, source=renderer_source)
def add_vline_static(fig, x, y, width, color):
return fig.line(x=x, y=y, line_width=width, color=color)
##############################################################################
# Read and store the data.
# - The wavelength values are encoded using an initial value $w_i$ and a fixed interval $w_d$ between data points. If total number of wavelength data points (```npoints```) is known, the full range can be recovered.
##############################################################################
with h5py.File(FILENAME, 'r') as f1:
y = []
if not FLAGS.no_stats:
means, stds = ([] for _ in range(2))
#print(f1['flux'].keys())
npoints = len(f1['flux']['flux_{}'.format('0'.zfill(NDIGITS))][0,:])
w_i = f1['meta']['wavelength'][0]
w_d = f1['meta']['wavelength'][1]
arr_x = np.arange(w_i, w_i+w_d*npoints, w_d)
for e in range(NEPOCHS):
estr = str(e).zfill(NDIGITS)
y.append( np.expand_dims(f1['fluxsamp']['flux_samp_'+estr][:NSPECTRA,:], axis=0) )
if not FLAGS.no_stats:
means.append( f1['meanssamp']['means_samp_'+estr][:] )
stds.append( f1['stdssamp']['stds_samp_'+estr][:] )
yreal = f1['flux']['flux_{}'.format('0'.zfill(NDIGITS))][:NSPECTRA,:]
if not FLAGS.no_stats:
meansreal = f1['means']['means_{}'.format('0'.zfill(NDIGITS))][:]
stdsreal = f1['stds']['stds_{}'.format('0'.zfill(NDIGITS))][:]
arr_y = np.vstack(tuple(x for x in y))
#print(arr_y.shape) NEPOCHS*NSPECTRA
if not FLAGS.no_stats:
arr_mean = np.vstack(tuple(x for x in means))
arr_std = np.vstack(tuple(x for x in stds))
##############################################################################
# Define further constants:
# - Strings to be used with *ColumnDataSource*s;
# - The minimum and maximum values of the histograms will be later useful for plotting.
##############################################################################
epochs = [x for x in range(NEPOCHS)]
s1, s2 = dict(), dict()
xs1, ys1 = 'wavelength', 'flux'
if not FLAGS.no_stats:
xs2, ys2 = ('means', 'stds'), ('meanscounts', 'stdscounts')
xse, yse = ('mean1x', 'mean2x'), ('mean1y', 'mean2y')
wstr = 'width1', 'width2'
def handle_histogram(arr):
h, edg = np.histogram(arr, NBINS)
c = (edg[:-1] + edg[1:])/2
return h, c
if not FLAGS.no_stats:
hstdmax, hmeanmax, cmeanmax, cstdmax = (-999 for _ in range(4))
hstdmin, hmeanmin, cmeanmin, cstdmin = (999 for _ in range(4))
hstdmaxreal, hmeanmaxreal, cmeanmaxreal, cstdmaxreal = (-999 for _ in range(4))
hstdminreal, hmeanminreal, cmeanminreal, cstdminreal = (999 for _ in range(4))
def get_min_max(data, dmax, dmin):
if np.max(data) > dmax:
dmax = np.max(data)
if np.min(data) < dmin:
dmin = np.min(data)
return dmax, dmin
for epoch in range(NEPOCHS):
s1name, s2name = 'source1_{}'.format(epoch), 'source2_{}'.format(epoch)
s1[s1name] = ColumnDataSource(data={xs1: arr_x.tolist()})
for i in range(NSPECTRA):
s1[s1name].data[ys1+str(i)] = (arr_y[epoch][i]+i*YSHIFT).tolist()
if not FLAGS.no_stats:
hmean, cmean = handle_histogram(arr_mean[epoch])
hstd, cstd = handle_histogram(arr_std[epoch])
s2[s2name] = ColumnDataSource(data={xs2[0]: cmean.tolist(),
ys2[0]: hmean.tolist(),
xs2[1]: cstd.tolist(),
ys2[1]: hstd.tolist(),
wstr[0]: [cmean[1]-cmean[0] for _ in range(len(cmean))],
wstr[1]: [cstd[1]-cstd[0] for _ in range(len(cstd))]})
if epoch==0:
s1[RENDERERS[0]] = ColumnDataSource(data={xs1: arr_x.tolist()})
for i in range(NSPECTRA):
s1[RENDERERS[0]].data[ys1+str(i)] = (arr_y[epoch][i]+i*YSHIFT).tolist()
if not FLAGS.no_stats:
initmeanx, initstdx = np.mean(cmean), np.mean(cstd)
s2[RENDERERS[1]] = ColumnDataSource(data={xs2[0]: cmean.tolist(),
ys2[0]: hmean.tolist(),
xs2[1]: cstd.tolist(),
ys2[1]: hstd.tolist(),
wstr[0]: [cmean[1]-cmean[0] for _ in range(len(cmean))],
wstr[1]: [cstd[1]-cstd[0] for _ in range(len(cstd))]})
if not FLAGS.no_stats:
hmeanmax, hmeanmin = get_min_max(hmean, hmeanmax, hmeanmin)
hstdmax, hstdmin = get_min_max(hstd, hstdmax, hstdmin)
cmeanmax, cmeanmin = get_min_max(cmean, cmeanmax, cmeanmin)
cstdmax, cstdmin = get_min_max(cstd, cstdmax, cstdmin)
if not FLAGS.no_stats:
hmeanreal, cmeanreal = handle_histogram(meansreal)
hstdreal, cstdreal = handle_histogram(stdsreal)
hmeanmaxreal, hmeanminreal = get_min_max(hmeanreal, hstdmaxreal, hstdminreal)
hstdmaxreal, hstdminreal = get_min_max(hstdreal, hstdmaxreal, hstdminreal)
cmeanmaxreal, cmeanminreal = get_min_max(cmeanreal, cmeanmaxreal, cmeanminreal)
cstdmaxreal, cstdminreal = get_min_max(cstdreal, cstdmaxreal, cstdminreal)
sreal1 = ColumnDataSource(data={xs1: arr_x.tolist()})
for i in range(NSPECTRA):
sreal1.data[ys1+str(i)] = (yreal[i]+i*YSHIFT).tolist()
if not FLAGS.no_stats:
sreal2 = ColumnDataSource(data={xs2[0]: cmeanreal.tolist(),
ys2[0]: hmeanreal.tolist(),
xs2[1]: cstdreal.tolist(),
ys2[1]: hstdreal.tolist(),
wstr[0]: [cmeanreal[1]-cmeanreal[0] for _ in range(len(cmeanreal))],
wstr[1]: [cstdreal[1]-cstdreal[0] for _ in range(len(cstdreal))]})
sreal2mean_mean = np.dot(cmeanreal,hstdreal)/np.sum(hstdreal)
sreal2mean_std = np.dot(cstdreal,hstdreal)/np.sum(hstdreal)
sextraname = 'sextra'
sextra = ColumnDataSource(data={xse[0]: [initmeanx,initmeanx], yse[0]: [hmeanmin, hmeanmax],
xse[1]: [initstdx,initstdx], yse[1]: [hstdmin, hstdmax]})
sextra_real = ColumnDataSource(data={xse[0]: [initmeanx,initmeanx], yse[0]: [hmeanmin, hmeanmax],
xse[1]: [initstdx,initstdx], yse[1]: [hstdmin, hstdmax]})
dict_sources_1 = dict(zip(epochs, ['source1_{}'.format(x) for x in epochs]))
js_sources_1 = str(dict_sources_1).replace("'", "")
if not FLAGS.no_stats:
dict_sources_2 = dict(zip(epochs, ['source2_{}'.format(x) for x in epochs]))
js_sources_2 = str(dict_sources_2).replace("'", "")
js_sources_e = sextraname
################################################################################################################
# Draw figures. The lines (```add_vline*```) start with dummy values so that they can be updated in the browser.
################################################################################################################
spectra_ylims = [-0.15,1.4]
#spectra_ylims = [-0.15,100.4]
spectra_yint = (spectra_ylims[0]-spectra_ylims[1])/15
spectraxlabel, spectraylabel = 'Wavelength [' + u'\u212b' + ']', 'Flux'
pfake = create_figure(WIDTH, HEIGHT, 'Generated QSO Spectra', ylims=spectra_ylims)
add_axis(pfake, xlabel=spectraxlabel, ylabel=spectraylabel,
ybounds=spectra_ylims, xinterval=w_d*npoints/5, yinterval=spectra_yint)
for i in range(NSPECTRA):
add_graph(pfake, xs1, ys1+str(i), s1[RENDERERS[0]], color=Category10[NSPECTRA][i])
preal = create_figure(WIDTH, HEIGHT, 'Real Spectra', ylims=spectra_ylims)
add_axis(preal, xlabel=spectraxlabel, ylabel=spectraylabel,
ybounds=spectra_ylims, xinterval=w_d*npoints/5, yinterval=spectra_yint)
for i in range(NSPECTRA):
add_graph(preal, xs1, ys1+str(i), sreal1, color=Category10[NSPECTRA][i])
if not FLAGS.no_stats:
p_mean = create_figure(int(WIDTH2), int(HEIGHT2), 'Generated Means',
xlims=[cmeanmin,cmeanmax], ylims=[hmeanmin,hmeanmax],
xaxislog=LOGAXIS)
add_axis(p_mean, xlabel=' ', ylabel='Counts',
xinterval=(cmeanmax-cmeanmin)/10, yinterval=hmeanmax/5)
add_bars(p_mean, xs2[0], ys2[0], wstr[0], s2[RENDERERS[1]], '#660044')
line = add_vline(p_mean, xse[0], yse[0], 3., sextra, '#008000')
add_legend(fig=p_mean, obj=line, label='Mean')
p_mean_real = create_figure(int(WIDTH2), int(HEIGHT2), 'Real Means',
xlims=[cmeanminreal,40],#xlims=[cmeanminreal,cmeanmaxreal],
ylims=[hmeanminreal,hmeanmaxreal],
xaxislog=LOGAXIS)
add_axis(p_mean_real, xlabel=' ', ylabel='Counts',
xinterval=(cmeanmaxreal-cmeanminreal)/40, yinterval=hmeanmaxreal/5)
add_bars(p_mean_real, xs2[0], ys2[0], wstr[0], sreal2, '#660044')
line = add_vline_static(p_mean_real, x=sreal2mean_mean, y=[hmeanminreal,hmeanmaxreal], width=3., color='#008000')
add_legend(fig=p_mean_real, obj=line, label='Mean')
p_std = create_figure(int(WIDTH2), int(HEIGHT2), 'Generated Standard Deviations',
xlims=[cstdmin,cstdmax], ylims=[hstdmin,hstdmax],
xaxislog=LOGAXIS)
add_axis(p_std, xlabel=' ', ylabel='Counts',
xinterval=(cstdmax-cstdmin)/10, yinterval=hstdmaxreal/5)
add_bars(p_std, xs2[1], ys2[1], wstr[1], s2[RENDERERS[1]], '#660044')
line = add_vline(p_std, xse[1], yse[1], 3., sextra, '#008000')
add_legend(fig=p_std, obj=line, label='Mean')
p_std_real = create_figure(int(WIDTH2), int(HEIGHT2), 'Real Standard Deviations',
xlims=[cstdminreal,20],#xlims=[cstdminreal,cstdmaxreal],
ylims=[hstdminreal,hstdmaxreal],
xaxislog=LOGAXIS)
add_axis(p_std_real, xlabel=' ', ylabel='Counts',
xinterval=(cstdmaxreal-cstdminreal)/40, yinterval=hstdmaxreal/5)
add_bars(p_std_real, xs2[1], ys2[1], wstr[1], sreal2, '#660044')
line_mean = add_vline_static(p_std_real, x=sreal2mean_std, y=[hstdminreal,hstdmaxreal], width=3., color='#008000')
add_legend(fig=p_std_real, obj=line_mean, label='Mean')
if FLAGS.no_stats:
code = """
var epoch = slider.value;
var s1 = {js_sources_1};
var new_s1 = s1[epoch].data;
s1_update.data = new_s1;
s1_update.change.emit();
""".format(js_sources_1=js_sources_1)
else:
code = """
var epoch = slider.value;
var s1 = {js_sources_1};
var s2 = {js_sources_2};
var se = {js_sources_e};
var new_s1 = s1[epoch].data;
s1_update.data = new_s1;
s1_update.change.emit();
var new_s2 = s2[epoch].data;
s2_update.data = new_s2;
s2_update.change.emit();
var d = s2_update.data;
var ncounts1_update = 0;
var mean1_update = 0.;
for (var i = 0; i<d['{means1}'].length; i++) {{
mean1_update += d['{meanscounts1}'][i] * d['{means1}'][i];
ncounts1_update += d['{meanscounts1}'][i];
}}
mean1_update /= ncounts1_update;
se.data['mean1x'][0] = mean1_update;
se.data['mean1x'][1] = mean1_update;
se.data['mean1y'][0] = 0;
se.data['mean1y'][1] = {vlinemax1};
var ncounts2_update = 0;
var mean2_update = 0.;
for (var i = 0; i<d['{means2}'].length; i++) {{
mean2_update += d['{meanscounts2}'][i] * d['{means2}'][i];
ncounts2_update += d['{meanscounts2}'][i];
}}
mean2_update /= ncounts2_update;
se.data['mean2x'][0] = mean2_update;
se.data['mean2x'][1] = mean2_update;
se.data['mean2y'][0] = 0;
se.data['mean2y'][1] = {vlinemax2};
se.change.emit();
""".format(js_sources_1=js_sources_1, js_sources_2=js_sources_2,
js_sources_e=js_sources_e,
vlinemax1=hmeanmax, vlinemax2=hstdmax,
wstr1=wstr[0],wstr2=wstr[1],
means1='means', means2='stds',
meanscounts1='meanscounts', meanscounts2='stdscounts')
callback_args = s1
if not FLAGS.no_stats:
callback_args.update(s2)
callback_args.update({sextraname: sextra})
callback = CustomJS(args=callback_args, code=code)
slider = Slider(start=0, end=NEPOCHS-1, value=0, step=1,
title='Epoch', width=int(WIDTH/2))
slider.js_on_change('value_throttled', callback)
callback.args['s1_update'] = s1[RENDERERS[0]]
if not FLAGS.no_stats:
callback.args['s2_update'] = s2[RENDERERS[1]]
callback.args['slider'] = slider
################################################################################################################
# Display
################################################################################################################
if FLAGS.no_stats:
display = layout([pfake, preal, slider])
else:
display = layout([[pfake, [p_mean, p_std]], [preal, [p_mean_real, p_std_real]], [slider]])
show(display)
#save(display) saves an independent html page