1 | #!/usr/bin/env python3 |
---|
2 | |
---|
3 | import sys |
---|
4 | import os |
---|
5 | import time |
---|
6 | import random |
---|
7 | import tempfile |
---|
8 | import argparse |
---|
9 | import socket |
---|
10 | import json |
---|
11 | |
---|
12 | import numpy |
---|
13 | import matplotlib.mlab as mlab |
---|
14 | import matplotlib.pyplot as plt |
---|
15 | |
---|
16 | |
---|
17 | VERSION = "{DEVELOPMENT}" |
---|
18 | if VERSION == "{DEVELOPMENT}": |
---|
19 | script_dir = '.' |
---|
20 | try: |
---|
21 | script_dir = os.path.dirname(os.path.realpath(__file__)) |
---|
22 | except: |
---|
23 | try: |
---|
24 | script_dir = os.path.dirname(os.path.abspath(sys.argv[0])) |
---|
25 | except: |
---|
26 | pass |
---|
27 | sys.path.append("%s/../lib" % script_dir) |
---|
28 | |
---|
29 | from nanownlib import * |
---|
30 | from nanownlib.stats import * |
---|
31 | import nanownlib.storage |
---|
32 | |
---|
33 | |
---|
34 | parser = argparse.ArgumentParser( |
---|
35 | description="") |
---|
36 | parser.add_argument('db_file', default=None, |
---|
37 | help='') |
---|
38 | options = parser.parse_args() |
---|
39 | db = nanownlib.storage.db(options.db_file) |
---|
40 | |
---|
41 | |
---|
42 | def differences(db, unusual_case, rtt_type='packet'): |
---|
43 | ret_val = [s['unusual_'+rtt_type]-s['other_'+rtt_type] for s in db.subseries('train', unusual_case)] |
---|
44 | ret_val += [s['unusual_'+rtt_type]-s['other_'+rtt_type] for s in db.subseries('test', unusual_case)] |
---|
45 | return ret_val |
---|
46 | |
---|
47 | def null_differences(db, unusual_case, rtt_type='packet'): |
---|
48 | ret_val = [s['unusual_'+rtt_type]-s['other_'+rtt_type] for s in db.subseries('train_null', unusual_case)] |
---|
49 | return ret_val |
---|
50 | |
---|
51 | |
---|
52 | def timeSeries(db, probe_type, unusual_case): |
---|
53 | cursor = db.conn.cursor() |
---|
54 | query=""" |
---|
55 | SELECT time_of_day,packet_rtt AS uc,(SELECT avg(packet_rtt) FROM probes,analysis |
---|
56 | WHERE analysis.probe_id=probes.id AND probes.test_case!=:unusual_case AND probes.type=:probe_type AND sample=u.sample) AS oc |
---|
57 | FROM (SELECT time_of_day,probes.sample,packet_rtt FROM probes,analysis |
---|
58 | WHERE analysis.probe_id=probes.id AND probes.test_case =:unusual_case AND probes.type=:probe_type) u |
---|
59 | """ |
---|
60 | |
---|
61 | params = {"probe_type":probe_type,"unusual_case":unusual_case} |
---|
62 | cursor.execute(query, params) |
---|
63 | for row in cursor: |
---|
64 | yield {'time_of_day':row['time_of_day'],unusual_case:row['uc'],'other_cases':row['oc']} |
---|
65 | #samples,derived,null_derived = parse_data(input1) |
---|
66 | |
---|
67 | #trust = trustValues(derived, sum) |
---|
68 | #weights = linearWeights(derived, trust, 0.25) |
---|
69 | #print('(test): %f' % weightedMean(derived,weights)) |
---|
70 | |
---|
71 | diffs = list(differences(db, 'long')) |
---|
72 | reported_diffs = list(differences(db, 'long', 'reported')) |
---|
73 | #shorts = [s['packet_rtt'] for s in samples.values() if s['test_case']=='short'] |
---|
74 | #longs = [s['packet_rtt'] for s in samples.values() if s['test_case']=='long'] |
---|
75 | |
---|
76 | short_overtime = [(sample['time_of_day'],sample['short']) for sample in timeSeries(db,'train','short')] |
---|
77 | long_overtime = [(sample['time_of_day'],sample['long']) for sample in timeSeries(db,'train','long')] |
---|
78 | diff_overtime = [(sample['time_of_day'],sample['long']-sample['other_cases']) for sample in timeSeries(db,'train','long')] |
---|
79 | short_overtime.sort() |
---|
80 | long_overtime.sort() |
---|
81 | diff_overtime.sort() |
---|
82 | |
---|
83 | print('packet_rtt diff median: %f' % statistics.median(diffs)) |
---|
84 | print('packet_rtt diff midhinge: %f' % midsummary(diffs)) |
---|
85 | print('packet_rtt diff trimean: %f' % trimean(diffs)) |
---|
86 | print('packet_rtt diff quadsummary: %f' % quadsummary(diffs)) |
---|
87 | print('packet_rtt diff ubersummary: %f' % ubersummary(diffs)) |
---|
88 | print('packet_rtt diff MAD: %f' % mad(diffs)) |
---|
89 | try: |
---|
90 | print('reported diff trimean: %f' % trimean(reported_diffs)) |
---|
91 | print('reported diff quadsummary: %f' % quadsummary(reported_diffs)) |
---|
92 | print('reported diff ubersummary: %f' % ubersummary(reported_diffs)) |
---|
93 | print('reported diff MAD: %f' % mad(reported_diffs)) |
---|
94 | |
---|
95 | import cProfile |
---|
96 | start = time.time() |
---|
97 | kresults = kfilter({},diffs) |
---|
98 | #print('packet_rtt diff kfilter: ', numpy.mean(kresults['est']), kresults['var']) |
---|
99 | print('packet_rtt diff kfilter: ', kresults['est'][-1], kresults['var'][-1]) |
---|
100 | kresults = kfilter({},reported_diffs) |
---|
101 | #print('reported diff kfilter: ', numpy.mean(kresults['est']), kresults['var'][-1]) |
---|
102 | print('reported diff kfilter: ', kresults['est'][-1], kresults['var'][-1]) |
---|
103 | print("kfilter time: %f" % (time.time()-start)) |
---|
104 | except: |
---|
105 | pass |
---|
106 | |
---|
107 | #print('tsval diff mean: %f' % numpy.mean(differences(db, 'long', 'tsval'))) |
---|
108 | #print('tsval null diff mean: %f' % numpy.mean(null_differences(db, 'long', 'tsval'))) |
---|
109 | #print('tsval diff weighted mean: %f' % tsvalwmean(db.subseries('train','long')+db.subseries('test','long'))) |
---|
110 | #print('tsval null diff weighted mean: %f' % tsvalwmean(db.subseries('train_null','long'))) |
---|
111 | |
---|
112 | |
---|
113 | def getTCPTSPrecision(): |
---|
114 | cursor = db.conn.cursor() |
---|
115 | query="""SELECT tcpts_mean FROM meta;""" |
---|
116 | cursor.execute(query) |
---|
117 | row = cursor.fetchone() |
---|
118 | if row: |
---|
119 | return row[0] |
---|
120 | return None |
---|
121 | |
---|
122 | |
---|
123 | def tsFilteredHistogram(): |
---|
124 | tcpts_precision = getTCPTSPrecision() |
---|
125 | |
---|
126 | num_bins = 500 |
---|
127 | all = db.subseries('train','long')+db.subseries('test','long') |
---|
128 | diffs = [s['unusual_packet']-s['other_packet'] for s in all] |
---|
129 | ts0_diffs = [s['unusual_packet']-s['other_packet'] for s in all if s['unusual_tsval']-s['other_tsval'] == 0] |
---|
130 | ts1_diffs = [s['unusual_packet']-s['other_packet'] for s in all if abs(s['unusual_tsval']-s['other_tsval']) > 0] |
---|
131 | ts2_diffs = [s['unusual_packet']-s['other_packet'] for s in all if abs(round((s['unusual_tsval']-s['other_tsval'])/tcpts_precision)) <= 1.0] |
---|
132 | |
---|
133 | ts_mode = statistics.mode([s['unusual_tsval'] for s in all]+[s['other_tsval'] for s in all]) |
---|
134 | ts_diff_mode = statistics.mode([s['unusual_tsval']-s['other_tsval'] for s in all]) |
---|
135 | ts_common_mode = [s['unusual_packet']-s['other_packet'] for s in all if s['unusual_tsval']<=ts_mode and s['other_tsval']<=ts_mode] |
---|
136 | ts_common_diff_mode = [s['unusual_packet']-s['other_packet'] for s in all if s['unusual_tsval']-s['other_tsval']==ts_diff_mode] |
---|
137 | |
---|
138 | print('packet_rtt diff quadsummary: %f' % quadsummary(diffs)) |
---|
139 | print('packet_rtt tsval diff=0 quadsummary: %f' % quadsummary(ts0_diffs)) |
---|
140 | print('packet_rtt tsval diff>0 quadsummary: %f' % quadsummary(ts1_diffs)) |
---|
141 | print('packet_rtt tsval diff<=1 quadsummary: %f' % quadsummary(ts2_diffs)) |
---|
142 | print('packet_rtt tsval mode quadsummary: %f' % quadsummary(ts_common_mode)) |
---|
143 | print(len(diffs), len(ts0_diffs)+len(ts1_diffs)) |
---|
144 | diffs.sort() |
---|
145 | cut_off_low = diffs[int(len(diffs)*0.005)] |
---|
146 | cut_off_high = diffs[int(len(diffs)*0.995)] |
---|
147 | |
---|
148 | plt.clf() |
---|
149 | # the histogram of the data |
---|
150 | n, bins, patches = plt.hist(diffs, num_bins, normed=0, color='black', histtype='step', alpha=0.8, |
---|
151 | range=(cut_off_low,cut_off_high), label='all') |
---|
152 | n, bins, patches = plt.hist(ts0_diffs, num_bins, normed=0, color='blue', histtype='step', alpha=0.8, |
---|
153 | range=(cut_off_low,cut_off_high), label='tsval diff=0') |
---|
154 | n, bins, patches = plt.hist(ts1_diffs, num_bins, normed=0, color='red', histtype='step', alpha=0.8, |
---|
155 | range=(cut_off_low,cut_off_high), label='tsval diff>0') |
---|
156 | n, bins, patches = plt.hist(ts2_diffs, num_bins, normed=0, color='orange', histtype='step', alpha=0.8, |
---|
157 | range=(cut_off_low,cut_off_high), label='tsval diff<=1') |
---|
158 | #n, bins, patches = plt.hist(ts_common_mode, num_bins, normed=0, color='green', histtype='step', alpha=0.8, |
---|
159 | # range=(cut_off_low,cut_off_high), label='tsval common mode') |
---|
160 | n, bins, patches = plt.hist(ts_common_diff_mode, num_bins, normed=0, color='green', histtype='step', alpha=0.8, |
---|
161 | range=(cut_off_low,cut_off_high), label='tsval common diff mode') |
---|
162 | plt.xlabel('RTT Difference') |
---|
163 | plt.ylabel('Probability') |
---|
164 | plt.title(r'Histogram - distribution of differences by tsval') |
---|
165 | |
---|
166 | # Tweak spacing to prevent clipping of ylabel |
---|
167 | plt.subplots_adjust(left=0.15) |
---|
168 | plt.legend() |
---|
169 | plt.show() |
---|
170 | #plt.savefig('paper/graphs/dists-vs-dist-of-diffs2.svg') |
---|
171 | |
---|
172 | tsFilteredHistogram() |
---|
173 | sys.exit(0) |
---|
174 | |
---|
175 | |
---|
176 | from pykalman import KalmanFilter |
---|
177 | #kf = KalmanFilter(transition_matrices = [[1, 1], [0, 1]], observation_matrices = [[0.1, 0.5], [-0.3, 0.0]]) |
---|
178 | kf = KalmanFilter(transition_matrices = [[1, 0], [0, 1]], n_dim_obs=2, observation_matrices = [[1.0, 0], [0, 1.0]]) |
---|
179 | |
---|
180 | #delta = 1e-5 |
---|
181 | #trans_cov = delta / (1 - delta) * np.eye(2) |
---|
182 | |
---|
183 | #kf = KalmanFilter(n_dim_obs=2, n_dim_state=2, |
---|
184 | # initial_state_mean=np.zeros(2), |
---|
185 | # initial_state_covariance=np.ones((2, 2)), |
---|
186 | # transition_matrices=np.eye(2), |
---|
187 | # observation_matrices=obs_mat, |
---|
188 | # observation_covariance=1.0, |
---|
189 | # transition_covariance=trans_cov) |
---|
190 | |
---|
191 | |
---|
192 | #measurements = numpy.asarray([[1,0], [0,0], [0,1]]) # 3 observations |
---|
193 | measurements = numpy.asarray([(s['unusual_packet'],s['other_packet']) for s in (db.subseries('train','long')+db.subseries('test','long'))]) |
---|
194 | kf = kf.em(measurements, n_iter=5) |
---|
195 | #(filtered_state_means, filtered_state_covariances) = kf.filter(measurements) |
---|
196 | #print("packet_rtt pykalman:", filtered_state_means[-1][0]-filtered_state_means[-1][1]) |
---|
197 | #print("packet_rtt pykalman:", filtered_state_means[-1]) |
---|
198 | |
---|
199 | (smoothed_state_means, smoothed_state_covariances) = kf.smooth(measurements) |
---|
200 | #up = numpy.mean([m[0] for m in smoothed_state_means]) |
---|
201 | #op = numpy.mean([m[1] for m in smoothed_state_means]) |
---|
202 | print("packet_rtt pykalman:", smoothed_state_means[-1], smoothed_state_means[-1][0]-smoothed_state_means[-1][1]) |
---|
203 | #print("packet_rtt pykalman:", up, op, up-op) |
---|
204 | |
---|
205 | |
---|
206 | #all_data = longs+shorts |
---|
207 | #all_data.sort() |
---|
208 | #cut_off_low = all_data[0] |
---|
209 | #cut_off_high = all_data[int(len(all_data)*0.997)] |
---|
210 | |
---|
211 | |
---|
212 | def plotSingleProbe(probe_id=None): |
---|
213 | if probe_id == None: |
---|
214 | cursor = db.conn.cursor() |
---|
215 | query="""SELECT probe_id FROM analysis WHERE suspect='' ORDER BY probe_id DESC limit 1 OFFSET 10""" |
---|
216 | cursor.execute(query) |
---|
217 | probe_id = cursor.fetchone()[0] |
---|
218 | |
---|
219 | cursor = db.conn.cursor() |
---|
220 | query="""SELECT observed,payload_len FROM packets WHERE probe_id=? AND sent=1""" |
---|
221 | cursor.execute(query, (probe_id,)) |
---|
222 | pkts = cursor.fetchall() |
---|
223 | sent_payload = [row[0] for row in pkts if row[1] != 0] |
---|
224 | sent_other = [row[0] for row in pkts if row[1] == 0] |
---|
225 | |
---|
226 | query="""SELECT observed,payload_len FROM packets WHERE probe_id=? AND sent=0""" |
---|
227 | cursor.execute(query, (probe_id,)) |
---|
228 | pkts = cursor.fetchall() |
---|
229 | rcvd_payload = [row[0] for row in pkts if row[1] != 0] |
---|
230 | rcvd_other = [row[0] for row in pkts if row[1] == 0] |
---|
231 | |
---|
232 | #query="""SELECT reported,time_of_day FROM probes WHERE id=?""" |
---|
233 | #cursor.execute(query, (probe_id,)) |
---|
234 | #reported,tod = cursor.fetchone() |
---|
235 | #userspace_times = [sent_times[0]-reported/3.0, sent_times[0]+reported] |
---|
236 | |
---|
237 | print("single probe counts:",len(sent_payload),len(sent_other),len(rcvd_payload),len(rcvd_other)) |
---|
238 | plt.clf() |
---|
239 | plt.title("Single HTTP Request - Packet Times") |
---|
240 | sp = plt.eventplot(sent_payload, colors=('red',), lineoffsets=8, linewidths=2, alpha=0.6,label='sent') |
---|
241 | so = plt.eventplot(sent_other, colors=('red',), lineoffsets=6, linewidths=2, alpha=0.6,label='sent') |
---|
242 | rp = plt.eventplot(rcvd_payload, colors=('blue',), lineoffsets=4, linewidths=2, alpha=0.6,label='received') |
---|
243 | ro = plt.eventplot(rcvd_other, colors=('blue',), lineoffsets=2, linewidths=2, alpha=0.6,label='received') |
---|
244 | #plt.legend((s,r), ('sent','received')) |
---|
245 | #plt.savefig('../img/http-packet-times.svg') |
---|
246 | plt.show() |
---|
247 | |
---|
248 | #plotSingleProbe() |
---|
249 | |
---|
250 | |
---|
251 | def graphTestResults(): |
---|
252 | plt.clf() |
---|
253 | plt.title("Test Results") |
---|
254 | plt.xlabel('sample size') |
---|
255 | plt.ylabel('error rate') |
---|
256 | legend = [] |
---|
257 | colors = ['red','blue','green','purple','orange','black','brown'] |
---|
258 | color_id = 0 |
---|
259 | |
---|
260 | cursor = db.conn.cursor() |
---|
261 | query = """ |
---|
262 | SELECT classifier FROM classifier_results GROUP BY classifier ORDER BY classifier; |
---|
263 | """ |
---|
264 | cursor.execute(query) |
---|
265 | classifiers = [] |
---|
266 | for c in cursor: |
---|
267 | classifiers.append(c[0]) |
---|
268 | |
---|
269 | for classifier in classifiers: |
---|
270 | query=""" |
---|
271 | SELECT params FROM classifier_results |
---|
272 | WHERE trial_type='test' |
---|
273 | AND classifier=:classifier |
---|
274 | AND (false_positives+false_negatives)/2.0 < 5.0 |
---|
275 | ORDER BY num_observations,(false_positives+false_negatives) |
---|
276 | LIMIT 1 |
---|
277 | """ |
---|
278 | cursor.execute(query, {'classifier':classifier}) |
---|
279 | row = cursor.fetchone() |
---|
280 | if row == None: |
---|
281 | query=""" |
---|
282 | SELECT params FROM classifier_results |
---|
283 | WHERE trial_type='test' and classifier=:classifier |
---|
284 | ORDER BY (false_positives+false_negatives),num_observations |
---|
285 | LIMIT 1 |
---|
286 | """ |
---|
287 | cursor.execute(query, {'classifier':classifier}) |
---|
288 | row = cursor.fetchone() |
---|
289 | if row == None: |
---|
290 | sys.stderr.write("WARN: couldn't find test results for classifier '%s'.\n" % classifier) |
---|
291 | continue |
---|
292 | |
---|
293 | best_params = row[0] |
---|
294 | query=""" |
---|
295 | SELECT num_observations,(false_positives+false_negatives)/2.0 FROM classifier_results |
---|
296 | WHERE trial_type='test' |
---|
297 | AND classifier=:classifier |
---|
298 | AND params=:params |
---|
299 | ORDER BY num_observations |
---|
300 | """ |
---|
301 | cursor.execute(query, {'classifier':classifier,'params':best_params}) |
---|
302 | |
---|
303 | num_obs = [] |
---|
304 | performance = [] |
---|
305 | for row in cursor: |
---|
306 | num_obs.append(row[0]) |
---|
307 | performance.append(row[1]) |
---|
308 | #print(num_obs,performance) |
---|
309 | path = plt.scatter(num_obs, performance, color=colors[color_id], s=4, alpha=0.8, linewidths=3.0) |
---|
310 | plt.plot(num_obs, performance, color=colors[color_id], alpha=0.8) |
---|
311 | legend.append((classifier,path)) |
---|
312 | color_id = (color_id+1) % len(colors) |
---|
313 | |
---|
314 | plt.legend([l[1] for l in legend], [l[0] for l in legend], scatterpoints=1, fontsize='xx-small') |
---|
315 | plt.show() |
---|
316 | |
---|
317 | graphTestResults() |
---|
318 | |
---|
319 | sys.exit(0) |
---|
320 | |
---|
321 | plt.clf() |
---|
322 | plt.title("Packet RTT over time") |
---|
323 | plt.xlabel('Time of Day') |
---|
324 | plt.ylabel('RTT') |
---|
325 | s = plt.scatter([t for t,rtt in short_overtime], [rtt for t,rtt in short_overtime], s=1, color='red', alpha=0.6) |
---|
326 | l = plt.scatter([t for t,rtt in long_overtime], [rtt for t,rtt in long_overtime], s=1, color='blue', alpha=0.6) |
---|
327 | d = plt.scatter([t for t,rtt in diff_overtime], [rtt for t,rtt in diff_overtime], s=1, color='purple', alpha=0.6) |
---|
328 | plt.legend((s,l,d), ('short','long','difference'), scatterpoints=1) |
---|
329 | #plt.savefig('paper/figures/comcast-powerboost1.png') |
---|
330 | plt.show() |
---|
331 | |
---|
332 | |
---|
333 | |
---|
334 | plt.clf() |
---|
335 | plt.title("Simple HTTP Request") |
---|
336 | plt.xlabel('Time of Day') |
---|
337 | plt.ylabel('') |
---|
338 | s = plt.scatter(sent_times, [2]*len(sent_times), s=3, color='red', alpha=0.9) |
---|
339 | r = plt.scatter(rcvd_times, [1]*len(rcvd_times), s=3, color='blue', alpha=0.9) |
---|
340 | plt.legend((s,r), ('sent','received'), scatterpoints=1) |
---|
341 | plt.show() |
---|
342 | |
---|
343 | sys.exit(0) |
---|
344 | short_overtime,long_overtime,diff_overtime = None,None,None |
---|
345 | |
---|
346 | |
---|
347 | num_bins = 300 |
---|
348 | reported_diffs.sort() |
---|
349 | cut_off_low = reported_diffs[int(len(diffs)*0.003)] |
---|
350 | cut_off_high = reported_diffs[int(len(diffs)*0.997)] |
---|
351 | |
---|
352 | plt.clf() |
---|
353 | # the histogram of the data |
---|
354 | n, bins, patches = plt.hist(reported_diffs, num_bins, normed=1, color='black', histtype='step', alpha=0.8, |
---|
355 | range=(cut_off_low,cut_off_high)) |
---|
356 | plt.xlabel('RTT Difference') |
---|
357 | plt.ylabel('Probability') |
---|
358 | plt.title(r'Histogram - distribution of differences') |
---|
359 | |
---|
360 | # Tweak spacing to prevent clipping of ylabel |
---|
361 | plt.subplots_adjust(left=0.15) |
---|
362 | #plt.legend() |
---|
363 | plt.show() |
---|
364 | #plt.savefig('paper/graphs/dists-vs-dist-of-diffs2.svg') |
---|
365 | |
---|
366 | |
---|
367 | |
---|
368 | |
---|
369 | num_bins = 300 |
---|
370 | diffs.sort() |
---|
371 | cut_off_low = diffs[int(len(diffs)*0.003)] |
---|
372 | cut_off_high = diffs[int(len(diffs)*0.997)] |
---|
373 | |
---|
374 | plt.clf() |
---|
375 | # the histogram of the data |
---|
376 | n, bins, patches = plt.hist(diffs, num_bins, normed=1, color='purple', histtype='step', alpha=0.8, |
---|
377 | range=(cut_off_low,cut_off_high)) |
---|
378 | plt.xlabel('RTT Difference') |
---|
379 | plt.ylabel('Probability') |
---|
380 | plt.title(r'Histogram - distribution of differences') |
---|
381 | |
---|
382 | # Tweak spacing to prevent clipping of ylabel |
---|
383 | plt.subplots_adjust(left=0.15) |
---|
384 | #plt.legend() |
---|
385 | plt.show() |
---|
386 | #plt.savefig('paper/graphs/dists-vs-dist-of-diffs2.svg') |
---|
387 | |
---|
388 | sys.exit(0) |
---|
389 | |
---|
390 | |
---|
391 | |
---|
392 | num_bins = 150 |
---|
393 | # the histogram of the data |
---|
394 | n, bins, patches = plt.hist((shorts,longs), num_bins, normed=1, label=['short', 'long'], color=['red','blue'], histtype='step', alpha=0.8, |
---|
395 | range=(cut_off_low,cut_off_high)) |
---|
396 | #n, bins, patches = plt.hist(shorts2+longs2, num_bins, normed=1, facecolor='blue', histtype='step', alpha=0.3) |
---|
397 | # add a 'best fit' line |
---|
398 | #y = mlab.normpdf(bins, mu, sigma) |
---|
399 | #plt.plot(bins, y, 'r--') |
---|
400 | plt.xlabel('packet_rtt') |
---|
401 | plt.ylabel('Probability') |
---|
402 | plt.title(r'Histogram - RTT short and long') |
---|
403 | |
---|
404 | # Tweak spacing to prevent clipping of ylabel |
---|
405 | plt.subplots_adjust(left=0.15) |
---|
406 | plt.legend() |
---|
407 | #plt.show() |
---|
408 | plt.savefig('paper/figures/comcast-powerboost2.svg') |
---|
409 | |
---|
410 | |
---|
411 | |
---|
412 | |
---|
413 | num_trials = 200 |
---|
414 | |
---|
415 | |
---|
416 | subsample_sizes = (50,150,300,500,700,1000,2000,3000,5000,7000,10000,15000,20000) |
---|
417 | estimator = functools.partial(boxTest, 0.07, 0.08) |
---|
418 | performance = [] |
---|
419 | for subsample_size in subsample_sizes: |
---|
420 | estimates = bootstrap(derived, subsample_size, num_trials, estimator) |
---|
421 | performance.append(100.0*len([e for e in estimates if e == 1])/num_trials) |
---|
422 | |
---|
423 | null_performance = [] |
---|
424 | for subsample_size in subsample_sizes: |
---|
425 | null_estimates = bootstrap(null_derived, subsample_size, num_trials, estimator) |
---|
426 | null_performance.append(100.0*len([e for e in null_estimates if e == 0])/num_trials) |
---|
427 | |
---|
428 | plt.clf() |
---|
429 | plt.title("boxTest bootstrap") |
---|
430 | plt.xlabel('sample size') |
---|
431 | plt.ylabel('performance') |
---|
432 | plt.scatter(subsample_sizes, performance, s=2, color='red', alpha=0.6) |
---|
433 | plt.scatter(subsample_sizes, null_performance, s=2, color='blue', alpha=0.6) |
---|
434 | plt.show() |
---|
435 | |
---|
436 | |
---|
437 | |
---|
438 | subsample_sizes = (50,150,300,400,500,700,1000,2000,3000,4000,5000,7000,10000) |
---|
439 | estimator = diffMedian |
---|
440 | performance = [] |
---|
441 | for subsample_size in subsample_sizes: |
---|
442 | estimates = bootstrap(derived, subsample_size, num_trials, estimator) |
---|
443 | performance.append(100.0*len([e for e in estimates if e > expected_mean*0.9 and e < expected_mean*1.1])/num_trials) |
---|
444 | |
---|
445 | plt.clf() |
---|
446 | plt.title("diff median bootstrap") |
---|
447 | plt.xlabel('sample size') |
---|
448 | plt.ylabel('performance') |
---|
449 | plt.scatter(subsample_sizes, performance, s=1, color='red', alpha=0.6) |
---|
450 | plt.show() |
---|
451 | |
---|
452 | |
---|
453 | |
---|
454 | |
---|
455 | subsample_sizes = (50,150,300,400,500,700,1000,2000,3000,4000,5000,7000,10000) |
---|
456 | weight_funcs = (linearWeights, prunedWeights) |
---|
457 | for wf in weight_funcs: |
---|
458 | estimator = functools.partial(estimateMean, hypotenuse, wf, 0.40) |
---|
459 | performance = [] |
---|
460 | for subsample_size in subsample_sizes: |
---|
461 | estimates = bootstrap(derived, subsample_size, num_trials, estimator) |
---|
462 | performance.append(100.0*len([e for e in estimates if e > expected_mean*0.9 and e < expected_mean*1.1])/num_trials) |
---|
463 | |
---|
464 | plt.clf() |
---|
465 | plt.title(repr(wf)) |
---|
466 | plt.xlabel('sample size') |
---|
467 | plt.ylabel('performance') |
---|
468 | plt.scatter(subsample_sizes, performance, s=1, color='red', alpha=0.6) |
---|
469 | plt.show() |
---|
470 | |
---|
471 | |
---|
472 | |
---|
473 | num_bins = 300 |
---|
474 | # the histogram of the data |
---|
475 | n, bins, patches = plt.hist((tsshorts,tslongs), num_bins, normed=1, label=['short', 'long'], color=['red','blue'], histtype='step', alpha=0.8) |
---|
476 | #n, bins, patches = plt.hist(shorts2+longs2, num_bins, normed=1, facecolor='blue', histtype='step', alpha=0.3) |
---|
477 | # add a 'best fit' line |
---|
478 | #y = mlab.normpdf(bins, mu, sigma) |
---|
479 | #plt.plot(bins, y, 'r--') |
---|
480 | plt.xlabel('packet_rtt') |
---|
481 | plt.ylabel('Probability') |
---|
482 | plt.title(r'Histogram - tsval_rtt short vs long') |
---|
483 | |
---|
484 | # Tweak spacing to prevent clipping of ylabel |
---|
485 | plt.subplots_adjust(left=0.15) |
---|
486 | plt.legend() |
---|
487 | plt.show() |
---|
488 | |
---|
489 | |
---|
490 | |
---|
491 | |
---|
492 | #### |
---|
493 | #trust_methods = [min,max,sum,difference,product] |
---|
494 | trust_methods = [sum,product,hypotenuse] |
---|
495 | colors = ['red','blue','green','purple','orange','black'] |
---|
496 | weight_methods = [prunedWeights, linearWeights] |
---|
497 | alphas = [i/100.0 for i in range(0,100,2)] |
---|
498 | |
---|
499 | |
---|
500 | |
---|
501 | |
---|
502 | plt.clf() |
---|
503 | plt.title(r'Trust Method Comparison - Linear') |
---|
504 | plt.xlabel('Alpha') |
---|
505 | plt.ylabel('Mean error') |
---|
506 | paths = [] |
---|
507 | for tm in trust_methods: |
---|
508 | trust = trustValues(derived, tm) |
---|
509 | series = [] |
---|
510 | for alpha in alphas: |
---|
511 | weights = linearWeights(derived, trust, alpha) |
---|
512 | series.append(weightedMean(derived, weights) - expected_mean) |
---|
513 | |
---|
514 | paths.append(plt.scatter(alphas, series, s=1, color=colors[len(paths)],alpha=0.6)) |
---|
515 | |
---|
516 | plt.legend(paths, [repr(tm) for tm in trust_methods], scatterpoints=1) |
---|
517 | plt.show() |
---|
518 | |
---|
519 | |
---|
520 | |
---|
521 | plt.clf() |
---|
522 | plt.title(r'Trust Method Comparison - Pruned') |
---|
523 | plt.xlabel('Alpha') |
---|
524 | plt.ylabel('Mean error') |
---|
525 | paths = [] |
---|
526 | for tm in trust_methods: |
---|
527 | trust = trustValues(derived, tm) |
---|
528 | series = [] |
---|
529 | for alpha in alphas: |
---|
530 | weights = prunedWeights(derived, trust, alpha) |
---|
531 | series.append(weightedMean(derived, weights) - expected_mean) |
---|
532 | |
---|
533 | paths.append(plt.scatter(alphas, series, s=1, color=colors[len(paths)],alpha=0.6)) |
---|
534 | |
---|
535 | plt.legend(paths, [repr(tm) for tm in trust_methods], scatterpoints=1) |
---|
536 | plt.show() |
---|
537 | |
---|
538 | |
---|
539 | sys.exit(0) |
---|
540 | |
---|
541 | plt.clf() |
---|
542 | plt.title(r'Trust Method Comparison - Inverted') |
---|
543 | plt.xlabel('Alpha') |
---|
544 | plt.ylabel('Mean error') |
---|
545 | paths = [] |
---|
546 | for tm in trust_methods: |
---|
547 | trust = trustValues(derived, tm) |
---|
548 | series = [] |
---|
549 | for alpha in alphas: |
---|
550 | weights = invertedWeights(derived, trust, alpha) |
---|
551 | series.append(weightedMean(derived, weights) - expected_mean) |
---|
552 | |
---|
553 | paths.append(plt.scatter(alphas, series, s=1, color=colors[len(paths)],alpha=0.6)) |
---|
554 | |
---|
555 | plt.legend(paths, [repr(tm) for tm in trust_methods], scatterpoints=1) |
---|
556 | plt.show() |
---|
557 | |
---|
558 | |
---|
559 | plt.clf() |
---|
560 | plt.title(r'Trust Method Comparison - Arctangent') |
---|
561 | plt.xlabel('Alpha') |
---|
562 | plt.ylabel('Mean error') |
---|
563 | paths = [] |
---|
564 | for tm in trust_methods: |
---|
565 | trust = trustValues(derived, tm) |
---|
566 | series = [] |
---|
567 | for alpha in alphas: |
---|
568 | weights = arctanWeights(derived, trust, alpha) |
---|
569 | series.append(weightedMean(derived, weights) - expected_mean) |
---|
570 | |
---|
571 | paths.append(plt.scatter(alphas, series, s=1, color=colors[len(paths)],alpha=0.6)) |
---|
572 | |
---|
573 | plt.legend(paths, [repr(tm) for tm in trust_methods], scatterpoints=1) |
---|
574 | plt.show() |
---|