def fruits_vs_chairs_L3(template=None):
"""
:param template: Template of features to sample from
:return: scope evaluation of fruits vs chair classification for dataset 1
"""
dataset = dataset1()
fruits_and_chairs_1 = lambda x: (x['category'] in frozenset(['fruits', 'chairs'])) and \
(x['obj'] in frozenset(dataset.obj_set1)) and \
(x['obj'] not in frozenset(Broken_objects))
# fruits_and_chairs_1 = {'obj': dataset.obj_set1, 'category': ['Fruits', 'Chairs']}
if template is None:
template = devthor_new_new_params.l3_params
eval_config = {
'npc_train': 40,
'npc_test': 40,
'npc_validate': 0,
'num_splits': 4,
'split_by': 'category',
'labelfunc': 'category',
'train_q': fruits_and_chairs_1,
'test_q': fruits_and_chairs_1,
'metric_screen': 'classifier',
'metric_kwargs': {'model_type': 'MCC2'}}
return scope.dp_sym_loss(template, dataset, eval_config)
def get_detection_windows(args):
(burnin, run_length, alpha_count_slow, alpha_count_fast,
alpha_mu_slow, alpha_mu_fast, buckets_slow, buckets_fast,
threshold) = args
np.random.seed((os.getpid() << 16) | (int(time.time()) & 0xFFFF))
rh_slow = create_rolling_histogram_class(
Bucket=create_bucket_class(
alpha_mu=alpha_mu_slow, alpha_count=alpha_count_slow),
target_buckets=buckets_slow)()
rh_fast = create_rolling_histogram_class(
Bucket=create_bucket_class(
alpha_mu=alpha_mu_fast, alpha_count=alpha_count_fast),
target_buckets=buckets_fast)()
#for val in gen_value([dist0], burnin):
# rh_slow.update(val)
# rh_fast.update(val)
data = []
#for i, val in enumerate(gen_value([dist1], run_length)):
for i, val in enumerate(datasets.dataset1()):
rh_slow.update(val)
rh_fast.update(val)
cdf_slow = rh_slow.get_CDF()
cdf_fast = rh_fast.get_CDF()
KS = get_KS(cdf_slow, cdf_fast)
if KS > threshold:
data.append(i)
return data
import numpy as np
import sys
import os
import sys
from roll import create_rolling_histogram_class, create_bucket_class
from roll import gen_value, get_KS
import time
import decay_equations
import datasets
from show_density import ergodic_chain, simple_histogram
dist0 = stats.norm(0, 1)
#dist1 = stats.norm(0, 1.5)
dist1 = datasets.dataset1()
def get_detection_windows(args):
(burnin, run_length, alpha_count_slow, alpha_count_fast,
alpha_mu_slow, alpha_mu_fast, buckets_slow, buckets_fast,
threshold) = args
np.random.seed((os.getpid() << 16) | (int(time.time()) & 0xFFFF))
rh_slow = create_rolling_histogram_class(
Bucket=create_bucket_class(
alpha_mu=alpha_mu_slow, alpha_count=alpha_count_slow),
target_buckets=buckets_slow)()
rh_fast = create_rolling_histogram_class(
Bucket=create_bucket_class(
alpha_mu=alpha_mu_fast, alpha_count=alpha_count_fast),
dists = [stats.expon(-2, 2), stats.norm(4, 2), stats.norm(-6, 1)]
upper = 10
for i in range(upper):
print "Generating data... {0} / {1}\r".format(i, upper),
sys.stdout.flush()
data = []
rh_slow = create_rolling_histogram_class(
Bucket=create_bucket_class(
alpha_mu=0.005, alpha_count=0.005),
target_buckets=200)()
rh_fast = create_rolling_histogram_class(
Bucket=create_bucket_class(
alpha_mu=0.005, alpha_count=0.01),
target_buckets=75)()
for val in datasets.dataset1():
rh_slow.update(val)
rh_fast.update(val)
cdf_long = rh_slow.get_CDF()
cdf_short = rh_fast.get_CDF()
data.append(calc_jaccard_distance(cdf_long, cdf_short))
# #for val in gen_value(dists[:2], size):
# for val in gen_value([stats.norm(0, 1)], size):
# rh_slow.update(val)
# rh_fast.update(val)
# cdf_long = rh_slow.get_CDF()
# cdf_short = rh_fast.get_CDF()
# data.append(calc_KS(cdf_long, cdf_short))
# #for val in gen_value(dists[2:3], size):
