def get_mean_exposures_per_pivot():
"Returns the number of exposures each pivot received."
cursor = connection.cursor()
cursor.execute("""
SELECT pivot, pivot_type, COUNT(*) as n_exposures
FROM drill_question
GROUP BY CONCAT(pivot, "|", pivot_type)
""")
data = cursor.fetchall()
word_c = []
kanji_c = []
combined_c = []
kanji_inc_dist = FreqDist()
for pivot, pivot_type, n_exposures in data:
combined_c.append(n_exposures)
if pivot_type == 'k':
kanji_c.append(n_exposures)
kanji_inc_dist.inc(pivot, n_exposures)
elif pivot_type == 'w':
word_c.append(n_exposures)
for kanji in scripts.unique_kanji(pivot):
kanji_inc_dist.inc(kanji, n_exposures)
else:
raise ValueError('unknown pivot type: %s' % pivot_type)
return [
('Words', mean(word_c)),
('Kanji', mean(kanji_c)),
('Combined', mean(combined_c)),
('Kanji combined', mean(kanji_inc_dist.values())),
]
def get_mean_exposures_per_pivot():
"Returns the number of exposures each pivot received."
cursor = connection.cursor()
cursor.execute("""
SELECT pivot, pivot_type, COUNT(*) as n_exposures
FROM drill_question
GROUP BY CONCAT(pivot, "|", pivot_type)
""")
data = cursor.fetchall()
word_c = []
kanji_c = []
combined_c = []
kanji_inc_dist = FreqDist()
for pivot, pivot_type, n_exposures in data:
combined_c.append(n_exposures)
if pivot_type == 'k':
kanji_c.append(n_exposures)
kanji_inc_dist.inc(pivot, n_exposures)
elif pivot_type == 'w':
word_c.append(n_exposures)
for kanji in scripts.unique_kanji(pivot):
kanji_inc_dist.inc(kanji, n_exposures)
else:
raise ValueError('unknown pivot type: %s' % pivot_type)
return [
('Words', mean(word_c)),
('Kanji', mean(kanji_c)),
('Combined', mean(combined_c)),
('Kanji combined', mean(kanji_inc_dist.values())),
]
def _calculate_pre_post_ratios(response_data):
"""
Returns the number of data which are correctly responded to on their first
presentation.
"""
response_data = [(pid, pt, i, ic)
for (i, (pid, pt, ic)) in enumerate(response_data)]
response_data.sort()
first_responses = []
last_responses = []
for (pivot_id, pivot_type), responses in groupby(response_data, lambda r:
(r[0], r[1])):
responses = list(responses)
if len(responses) < 2:
continue
first_responses.append(responses[0][3])
last_responses.append(responses[-1][3])
if not first_responses:
return None, None
return (
mean(first_responses),
mean(last_responses),
)
def _calculate_pre_post_ratios(response_data):
"""
Returns the number of data which are correctly responded to on their first
presentation.
"""
response_data = [(pid, pt, i, ic) for (i, (pid, pt, ic)) in
enumerate(response_data)]
response_data.sort()
first_responses = []
last_responses = []
for (pivot_id, pivot_type), responses in groupby(response_data,
lambda r: (r[0], r[1])):
responses = list(responses)
if len(responses) < 2:
continue
first_responses.append(responses[0][3])
last_responses.append(responses[-1][3])
if not first_responses:
return None, None
return (
mean(first_responses),
mean(last_responses),
)
def get_global_rater_stats():
cursor = connection.cursor()
cursor.execute("""
SELECT id, username
FROM auth_user
""")
id_to_username = dict(cursor.fetchall())
cursor.execute("""
SELECT q_time.user_id, q.pivot, q.pivot_type, q_time.is_correct
FROM (
SELECT mco.question_id, mco.is_correct, ot.user_id, ot.timestamp
FROM (
SELECT mcr.option_id, dr.user_id, dr.timestamp
FROM drill_response AS dr
INNER JOIN drill_multiplechoiceresponse AS mcr
ON mcr.response_ptr_id = dr.id
) AS ot
INNER JOIN drill_multiplechoiceoption AS mco
ON mco.id = ot.option_id
) AS q_time
INNER JOIN drill_question AS q
ON q.id = q_time.question_id
ORDER BY user_id ASC, q_time.timestamp ASC
""")
results = []
ignore_users = _get_user_ignore_set()
for user_id, rows in groupby(cursor.fetchall(), lambda r: r[0]):
if user_id in ignore_users:
continue
rows = [(p, pt, c) for (_u, p, pt, c) in rows] # discard user_id
user_data = {
'user_id': user_id,
'username': id_to_username[user_id]
}
user_data['n_responses'] = len(rows)
user_data['n_tests'] = drill_models.TestSet.objects.filter(
user__id=user_id).exclude(end_time=None).count()
user_data['mean_accuracy'] = mean(r[2] for r in rows)
user_data['n_errors'] = _seq_len(r for r in rows if r[2])
pre_ratio, post_ratio = _calculate_pre_post_ratios(rows)
user_data['pre_ratio'] = pre_ratio
user_data['post_ratio'] = post_ratio
user_data['state_machine'] = _calculated_item_change(rows)
if pre_ratio and post_ratio:
user_data['pre_post_diff'] = post_ratio - pre_ratio
else:
user_data['pre_post_diff'] = None
results.append(user_data)
return results
def get_rater_stats(rater):
responses = drill_models.MultipleChoiceOption.objects.filter(
multiplechoiceresponse__user=rater).values('is_correct')
mean_accuracy = mean((r['is_correct'] and 1 or 0) for r in responses)
return {
'n_responses':
drill_models.Response.objects.filter(user=rater).count(),
'n_tests': drill_models.TestSet.objects.filter(user=rater).count(),
'mean_accuracy': mean_accuracy,
}
def get_global_rater_stats():
cursor = connection.cursor()
cursor.execute("""
SELECT id, username
FROM auth_user
""")
id_to_username = dict(cursor.fetchall())
cursor.execute("""
SELECT q_time.user_id, q.pivot, q.pivot_type, q_time.is_correct
FROM (
SELECT mco.question_id, mco.is_correct, ot.user_id, ot.timestamp
FROM (
SELECT mcr.option_id, dr.user_id, dr.timestamp
FROM drill_response AS dr
INNER JOIN drill_multiplechoiceresponse AS mcr
ON mcr.response_ptr_id = dr.id
) AS ot
INNER JOIN drill_multiplechoiceoption AS mco
ON mco.id = ot.option_id
) AS q_time
INNER JOIN drill_question AS q
ON q.id = q_time.question_id
ORDER BY user_id ASC, q_time.timestamp ASC
""")
results = []
ignore_users = _get_user_ignore_set()
for user_id, rows in groupby(cursor.fetchall(), lambda r: r[0]):
if user_id in ignore_users:
continue
rows = [(p, pt, c) for (_u, p, pt, c) in rows] # discard user_id
user_data = {'user_id': user_id, 'username': id_to_username[user_id]}
user_data['n_responses'] = len(rows)
user_data['n_tests'] = drill_models.TestSet.objects.filter(
user__id=user_id).exclude(end_time=None).count()
user_data['mean_accuracy'] = mean(r[2] for r in rows)
user_data['n_errors'] = _seq_len(r for r in rows if r[2])
pre_ratio, post_ratio = _calculate_pre_post_ratios(rows)
user_data['pre_ratio'] = pre_ratio
user_data['post_ratio'] = post_ratio
user_data['state_machine'] = _calculated_item_change(rows)
if pre_ratio and post_ratio:
user_data['pre_post_diff'] = post_ratio - pre_ratio
else:
user_data['pre_post_diff'] = None
results.append(user_data)
return results
def get_rater_stats(rater):
responses = drill_models.MultipleChoiceOption.objects.filter(
multiplechoiceresponse__user=rater).values('is_correct')
mean_accuracy = mean((r['is_correct'] and 1 or 0) for r in responses)
return {
'n_responses': drill_models.Response.objects.filter(
user=rater).count(),
'n_tests': drill_models.TestSet.objects.filter(
user=rater).count(),
'mean_accuracy': mean_accuracy,
}
def sample_by_centroid(cls, locations, potential_locations,
use_max=False):
mean_x = mean(x for (x, y) in locations)
mean_y = mean(y for (x, y) in locations)
parts = []
n_locations = len(potential_locations)
for loc in potential_locations:
x, y = loc
dist = sqrt((x - mean_x)**2 + (y - mean_y)**2)
if dist > 0:
score = 1 / dist**2
else:
score = 1
parts.append((loc, score))
dist = prob.ProbDist(parts)
if use_max:
p, l = max((p, l) for (l, p) in dist.iteritems())
return l
return dist.sample()
if data[width][nodes][ch].has_key('prim'):
data[width][nodes][ch]['prim'].append(prim)
else:
data[width][nodes][ch]['prim'] = [prim]
if data[width][nodes][ch].has_key('rcs'):
data[width][nodes][ch]['rcs'].append(rcs)
else:
data[width][nodes][ch]['rcs'] = [rcs]
for width in sorted(data.keys()):
for nodes in sorted(data[width].keys()):
for channels in sorted(data[width][nodes].keys()):
rd = data[width][nodes][channels]
rd['dijkstra'] = simplestats.mean(rd['dijkstra'])
rd['prim'] = simplestats.mean(rd['prim'])
rd['rcs'] = simplestats.mean(rd['rcs'])
# Scenario 1: vary # of channels in {1, 2, 3, 4, 5}
d = data[50000.0][25]
dij = []
pri = []
rcs = []
for channels in sorted(d.keys()):
dij.append(str(d[channels]['dijkstra'] / 1000000))
pri.append(str(d[channels]['prim'] / 1000000))
rcs.append(str(d[channels]['rcs'] / 1000000))
print(
"""
def _accumulate_plugin_errors(raw_data):
data = []
for label, scores in itertools.groupby(raw_data, lambda x: x[0]):
data.append((label, mean(v for (l, v) in scores)))
return data
# for j in data[i].keys():
# for k in data[i][j].keys():
# for l in data[i][j][k].keys():
# for m in data[i][j][k][l].keys():
# print "W: %d N: %d C: %d AP: %d V: %s #: %d" % (i, j, k, l, m, len(data[i][j][k][l][m]))
# Scenario 1: vary size of the network
mst = []
mst2 = []
msto = []
mst2o = []
idxs = []
for width in sorted(data.keys()):
mst.append(str(simplestats.mean(data[width][MUSERS][MCHAN][0]['mst'])))
mst2.append(str(simplestats.mean(data[width][MUSERS][MCHAN][0]['mst2'])))
msto.append(str(simplestats.mean(data[width][MUSERS][MCHAN][0]['mst-apx'])))
mst2o.append(str(simplestats.mean(data[width][MUSERS][MCHAN][0]['mst2-apx'])))
idxs.append(int(width/1000))
print("""
figure(1);
set(1, \"defaulttextfontname\", \"Times-Roman\");
set(1, \"defaultaxesfontname\", \"Times-Roman\");
set(1, \"defaulttextfontsize\", 19);
set(1, \"defaultaxesfontsize\", 19);
X = %(idx)s;
MST = [ %(mst)s ];
MST2 = [ %(mst2)s ];
if data[width][theta][relays][subs][mq][ch].has_key('r2'):
data[width][theta][relays][subs][mq][ch]['r2'].append(r2)
else:
data[width][theta][relays][subs][mq][ch]['r2'] = [r2]
for width in sorted(data.keys()):
for theta in sorted(data[width].keys()):
for relay in sorted(data[width][theta].keys()):
for subs in sorted(data[width][theta][relay].keys()):
for mq in sorted(data[width][theta][relay][subs].keys()):
for ch in sorted(data[width][theta][relay][subs][mq].keys()):
rd = data[width][theta][relay][subs][mq][ch]
lilp = len(rd['ilp'])
lr1 = len(rd['r1'])
lr2 = len(rd['r2'])
rd['ilp'] = simplestats.mean(rd['ilp'])
rd['r1'] = simplestats.mean(rd['r1'])
rd['r2'] = simplestats.mean(rd['r2'])
# Scenario 1: vary # of subscribers in {20, 30, 40, 50, 60}
d = data[40000.0][40][4]
ilp = []
r1 = []
r2 = []
for subs in sorted(d.keys()):
ilp.append(str(d[subs][40000.0][4]['ilp'] / 1000000))
r1.append(str(d[subs][40000.0][4]['r1'] / 1000000))
r2.append(str(d[subs][40000.0][4]['r2'] / 1000000))
print("""
figure(1);
if data[width][nodes][ch].has_key('rcsdpcs'):
data[width][nodes][ch]['rcsdpcs'].append(rcsdpcs)
else:
data[width][nodes][ch]['rcsdpcs'] = [rcsdpcs]
if data[width][nodes][ch].has_key('rcs'):
data[width][nodes][ch]['rcs'].append(rcs)
else:
data[width][nodes][ch]['rcs'] = [rcs]
for width in sorted(data.keys()):
for nodes in sorted(data[width].keys()):
for channels in sorted(data[width][nodes].keys()):
rd = data[width][nodes][channels]
rd['spdpcs'] = simplestats.mean(rd['spdpcs'])
rd['spgdy'] = simplestats.mean(rd['spgdy'])
rd['btldpcs'] = simplestats.mean(rd['btldpcs'])
rd['btlgdy'] = simplestats.mean(rd['btlgdy'])
rd['rcsdpcs'] = simplestats.mean(rd['rcsdpcs'])
rd['rcs'] = simplestats.mean(rd['rcs'])
# Scenario 1: vary # of channels in {1, 2, 3, 4, 5}
d = data[50000.0][25]
spdpcs = []
spgdy = []
btldpcs = []
btlgdy = []
rcsdpcs = []
rcs = []
idx = [9, 16, 25, 36, 49];
def _accumulate_plugin_errors(raw_data):
data = []
for label, scores in itertools.groupby(raw_data, lambda x: x[0]):
data.append((label, mean(v for (l, v) in scores)))
return data
if data[width][nodes][ch].has_key('prim'):
data[width][nodes][ch]['prim'].append(prim)
else:
data[width][nodes][ch]['prim'] = [prim]
if data[width][nodes][ch].has_key('rcs'):
data[width][nodes][ch]['rcs'].append(rcs)
else:
data[width][nodes][ch]['rcs'] = [rcs]
for width in sorted(data.keys()):
for nodes in sorted(data[width].keys()):
for channels in sorted(data[width][nodes].keys()):
rd = data[width][nodes][channels]
rd['dijkstra'] = simplestats.mean(rd['dijkstra'])
rd['prim'] = simplestats.mean(rd['prim'])
rd['rcs'] = simplestats.mean(rd['rcs'])
# Scenario 1: vary # of channels in {1, 2, 3, 4, 5}
d = data[50000.0][25]
dij = []
pri = []
rcs = []
for channels in sorted(d.keys()):
dij.append(str(d[channels]['dijkstra'] / 1000000))
pri.append(str(d[channels]['prim'] / 1000000))
rcs.append(str(d[channels]['rcs'] / 1000000))
print("""
figure(1);